Skip to content

GitLab

Commit 41b18fd8 authored by zhe chen's avatar zhe chen
Browse files

Use pre-commit to reformat code 


Use pre-commit to reformat code
parent ff20ea39
Show whitespace changes
Inline Side-by-side
Showing with 362 additions and 421 deletions
autonomous_driving/Online-HD-Map-Construction/src/models/heads/polyline_generator.py
View file @ 41b18fd8
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
from mmcv.runner import force_fp32
from mmdet.models import HEADS
from torch.distributions.categorical import Categorical
from mmdet.models import HEADS
from .detgen_utils.causal_trans import (CausalTransformerDecoder,
CausalTransformerDecoderLayer)
from .detgen_utils.utils import (dequantize_verts, generate_square_subsequent_mask,
quantize_verts, top_k_logits, top_p_logits)
from mmcv.runner import force_fp32, auto_fp16
from .detgen_utils.utils import (generate_square_subsequent_mask, top_k_logits,
top_p_logits)
@HEADS.register_module(force=True)
class PolylineGenerator(nn.Module):
......@@ -63,7 +63,7 @@ class PolylineGenerator(nn.Module):
self.fp16_enabled = False
self.coord_dim = coord_dim # if we use xyz else 2 when we use xy
self.kp_coord_dim = coord_dim if coord_dim==2 else 2 # XXX
self.kp_coord_dim = coord_dim if coord_dim == 2 else 2 # XXX
self.register_buffer('canvas_size', torch.tensor(canvas_size))
# initialize the model
......@@ -126,7 +126,7 @@ class PolylineGenerator(nn.Module):
# Discrete vertex value embeddings
vert_embeddings = self.vertex_embed(bbox)
return vert_embeddings + (bbox_embedding+coord_embeddings)[None]
return vert_embeddings + (bbox_embedding + coord_embeddings)[None]
def _prepare_context(self, batch, context):
"""Prepare class label and vertex context."""
......@@ -189,7 +189,7 @@ class PolylineGenerator(nn.Module):
# Aggregate embeddings
embeddings = vert_embeddings + \
(coord_embeddings+pos_embeddings)[None]
(coord_embeddings + pos_embeddings)[None]
embeddings = torch.cat([condition_embedding, embeddings], dim=1)
return embeddings
......@@ -218,24 +218,22 @@ class PolylineGenerator(nn.Module):
sizes = [size, polyline_length.max()]
polyline_logits = []
for c_idx, size in zip([c1,c2], sizes):
new_batch = assign_batch(batch,c_idx, size)
_poly_logits = self._forward_train(new_batch,context,**kwargs)
for c_idx, size in zip([c1, c2], sizes):
new_batch = assign_batch(batch, c_idx, size)
_poly_logits = self._forward_train(new_batch, context, **kwargs)
polyline_logits.append(_poly_logits)
# maybe imporve the speed
for i, (_poly_logits, size) in enumerate(zip(polyline_logits, sizes)):
if size < sizes[1]:
_poly_logits = F.pad(_poly_logits, (0,0,0,sizes[1]-size), "constant", 0)
_poly_logits = F.pad(_poly_logits, (0, 0, 0, sizes[1] - size), 'constant', 0)
polyline_logits[i] = _poly_logits
polyline_logits = torch.cat(polyline_logits,0)
polyline_logits = torch.cat(polyline_logits, 0)
polyline_logits = polyline_logits[revert_idx]
cat_dist = Categorical(logits=polyline_logits)
return {'polylines':cat_dist}
return {'polylines': cat_dist}
def forward_train(self, batch: dict, context: dict, **kwargs):
"""
......@@ -247,7 +245,7 @@ class PolylineGenerator(nn.Module):
if False:
polyline_logits = self._forward_train(batch, context, **kwargs)
cat_dist = Categorical(logits=polyline_logits)
return {'polylines':cat_dist}
return {'polylines': cat_dist}
else:
return self.sperate_forward(batch, context, **kwargs)
......@@ -271,7 +269,7 @@ class PolylineGenerator(nn.Module):
return logits
@force_fp32(apply_to=('global_context_embedding','sequential_context_embeddings','cache'))
@force_fp32(apply_to=('global_context_embedding', 'sequential_context_embeddings', 'cache'))
def body(self,
seqs,
global_context_embedding=None,
......@@ -314,28 +312,27 @@ class PolylineGenerator(nn.Module):
decoder_outputs = decoder_outputs.transpose(0, 1)
# since we only need the predict seq
decoder_outputs = decoder_outputs[:, condition_len-1:]
decoder_outputs = decoder_outputs[:, condition_len - 1:]
# Get logits and optionally process for sampling
logits = self._project_to_logits(decoder_outputs)
# y mask
_vert_mask = torch.arange(logits.shape[-1], device=logits.device)
vertices_mask_y = (_vert_mask < self.canvas_size[1]+1)
vertices_mask_y = (_vert_mask < self.canvas_size[1] + 1)
vertices_mask_y[0] = False # y position doesn't have stop sign
logits[:, 1::self.coord_dim] = logits[:, 1::self.coord_dim] * \
vertices_mask_y - ~vertices_mask_y*1e9
vertices_mask_y - ~vertices_mask_y * 1e9
if self.coord_dim > 2:
# z mask
_vert_mask = torch.arange(logits.shape[-1], device=logits.device)
vertices_mask_z = (_vert_mask < self.canvas_size[2]+1)
vertices_mask_z = (_vert_mask < self.canvas_size[2] + 1)
vertices_mask_z[0] = False # y position doesn't have stop sign
logits[:, 2::self.coord_dim] = logits[:, 2::self.coord_dim] * \
vertices_mask_z - ~vertices_mask_z*1e9
vertices_mask_z - ~vertices_mask_z * 1e9
logits = logits/temperature
logits = logits / temperature
logits = top_k_logits(logits, top_k)
logits = top_p_logits(logits, top_p)
if return_logits:
......@@ -352,7 +349,7 @@ class PolylineGenerator(nn.Module):
mask = gt['polyline_masks']
loss = -torch.sum(
pred['polylines'].log_prob(gt['polylines']) * mask * weight)/weight.sum()
pred['polylines'].log_prob(gt['polylines']) * mask * weight) / weight.sum()
return {'seq': loss}
......@@ -395,16 +392,15 @@ class PolylineGenerator(nn.Module):
samples = torch.empty(
[batch_size, 0], dtype=torch.int32, device=device)
max_sample_length = max_sample_length or self.max_seq_length
seq_len = max_sample_length*self.coord_dim+1
seq_len = max_sample_length * self.coord_dim + 1
cache = None
decoded_tokens = \
torch.zeros((batch_size,seq_len),
device=device,dtype=torch.long)
torch.zeros((batch_size, seq_len),
device=device, dtype=torch.long)
remain_idx = torch.arange(batch_size, device=device)
for i in range(seq_len):
# While-loop body for autoregression calculation.
pred_dist, cache = self.body(
samples,
......@@ -417,22 +413,22 @@ class PolylineGenerator(nn.Module):
is_training=False)
samples = pred_dist.sample()
decoded_tokens[remain_idx,i] = samples[:,-1]
decoded_tokens[remain_idx, i] = samples[:, -1]
# Stopping conditions for autoregressive calculation.
if not (decoded_tokens[:,:i+1] != 0).all(-1).any():
if not (decoded_tokens[:, :i + 1] != 0).all(-1).any():
break
# update state, check the new position is zero.
valid_idx = (samples[:,-1] != 0).nonzero(as_tuple=True)[0]
valid_idx = (samples[:, -1] != 0).nonzero(as_tuple=True)[0]
remain_idx = remain_idx[valid_idx]
cache = cache[:,:,valid_idx]
cache = cache[:, :, valid_idx]
global_context = global_context[valid_idx]
seq_context = seq_context[valid_idx]
samples = samples[valid_idx]
# decoded_tokens = torch.cat(decoded_tokens,dim=1)
decoded_tokens = decoded_tokens[:,:i+1]
decoded_tokens = decoded_tokens[:, :i + 1]
outputs = self.post_process(decoded_tokens, seq_len,
device, only_return_complete)
......@@ -455,7 +451,7 @@ class PolylineGenerator(nn.Module):
_polyline_mask = torch.arange(sample_seq_length)[None].to(device)
# Get largest stopping point for incomplete samples.
valid_polyline_len = torch.full_like(polyline[:,0], sample_seq_length)
valid_polyline_len = torch.full_like(polyline[:, 0], sample_seq_length)
zero_inds = (polyline == 0).type(torch.int32).argmax(-1)
# Real length
......@@ -463,7 +459,7 @@ class PolylineGenerator(nn.Module):
polyline_mask = _polyline_mask < valid_polyline_len[:, None]
# Mask faces beyond stopping token with zeros
polyline = polyline*polyline_mask
polyline = polyline * polyline_mask
# Pad to maximum size with zeros
pad_size = max_seq_len - sample_seq_length
......@@ -486,25 +482,25 @@ class PolylineGenerator(nn.Module):
}
return outputs
def find_best_sperate_plan(idx,array):
def find_best_sperate_plan(idx, array):
h = array[-1] - array[idx]
w = idx
cost = h*w
cost = h * w
return cost
def get_chunk_idx(polyline_length):
_polyline_length, polyline_length_idx = torch.sort(polyline_length)
costs = []
for i in range(len(_polyline_length)):
cost = find_best_sperate_plan(i,_polyline_length)
cost = find_best_sperate_plan(i, _polyline_length)
costs.append(cost)
seperate_point = torch.stack(costs).argmax()
chunk1 = polyline_length_idx[:seperate_point+1]
chunk2 = polyline_length_idx[seperate_point+1:]
chunk1 = polyline_length_idx[:seperate_point + 1]
chunk2 = polyline_length_idx[seperate_point + 1:]
revert_idx = torch.argsort(polyline_length_idx)
......@@ -517,9 +513,9 @@ def assign_bev(feat, idx):
def assign_batch(batch, idx, size):
new_batch = {}
for k,v in batch.items():
for k, v in batch.items():
new_batch[k] = v[idx]
if new_batch[k].ndim > 1:
new_batch[k] = new_batch[k][:,:size]
new_batch[k] = new_batch[k][:, :size]
return new_batch
autonomous_driving/Online-HD-Map-Construction/src/models/losses/detr_loss.py
View file @ 41b18fd8
import mmcv
import torch
from mmdet.models.builder import LOSSES
from mmdet.models.losses.utils import weighted_loss
from torch import nn as nn
from torch.nn import functional as F
from mmdet.models.losses import l1_loss
from mmdet.models.losses.utils import weighted_loss
import mmcv
from mmdet.models.builder import LOSSES
@mmcv.jit(derivate=True, coderize=True)
......@@ -77,7 +75,7 @@ class LinesLoss(nn.Module):
loss = smooth_l1_loss(
pred, target, weight, reduction=reduction, avg_factor=avg_factor, beta=self.beta)
return loss*self.loss_weight
return loss * self.loss_weight
@mmcv.jit(derivate=True, coderize=True)
......@@ -123,7 +121,8 @@ class MasksLoss(nn.Module):
loss = bce(pred, target, weight, reduction=reduction,
avg_factor=avg_factor)
return loss*self.loss_weight
return loss * self.loss_weight
@mmcv.jit(derivate=True, coderize=True)
@weighted_loss
......@@ -167,4 +166,4 @@ class LenLoss(nn.Module):
loss = ce(pred, target, weight, reduction=reduction,
avg_factor=avg_factor)
return loss*self.loss_weight
\ No newline at end of file
return loss * self.loss_weight
autonomous_driving/Online-HD-Map-Construction/src/models/mapers/base_mapper.py
View file @ 41b18fd8
from abc import ABCMeta, abstractmethod
from abc import ABCMeta
import torch.nn as nn
from mmcv.runner import auto_fp16
from mmcv.utils import print_log
from mmdet.utils import get_root_logger
from mmdet3d.models.builder import DETECTORS
from mmdet.utils import get_root_logger
MAPPERS = DETECTORS
class BaseMapper(nn.Module, metaclass=ABCMeta):
"""Base class for mappers."""
......@@ -40,7 +39,7 @@ class BaseMapper(nn.Module, metaclass=ABCMeta):
return ((hasattr(self, 'roi_head') and self.roi_head.with_mask)
or (hasattr(self, 'mask_head') and self.mask_head is not None))
#@abstractmethod
# @abstractmethod
def extract_feat(self, imgs):
"""Extract features from images."""
pass
......@@ -48,11 +47,11 @@ class BaseMapper(nn.Module, metaclass=ABCMeta):
def forward_train(self, *args, **kwargs):
pass
#@abstractmethod
# @abstractmethod
def simple_test(self, img, img_metas, **kwargs):
pass
#@abstractmethod
# @abstractmethod
def aug_test(self, imgs, img_metas, **kwargs):
"""Test function with test time augmentation."""
pass
......
autonomous_driving/Online-HD-Map-Construction/src/models/mapers/vectormapnet.py
View file @ 41b18fd8
import mmcv
import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
from mmdet3d.models.builder import build_backbone, build_head, build_neck
from torch.nn.utils.rnn import pad_sequence
from torchvision.models.resnet import resnet18, resnet50
from torchvision.models.resnet import resnet18
from mmdet3d.models.builder import (build_backbone, build_head,
build_neck)
from .base_mapper import BaseMapper, MAPPERS
from .base_mapper import MAPPERS, BaseMapper
@MAPPERS.register_module()
......@@ -31,8 +27,7 @@ class VectorMapNet(BaseMapper):
model_name=None, **kwargs):
super(VectorMapNet, self).__init__()
#Attribute
# Attribute
self.model_name = model_name
self.last_epoch = None
self.only_det = only_det
......@@ -59,11 +54,10 @@ class VectorMapNet(BaseMapper):
)
# BEV
if hasattr(self.backbone,'bev_w'):
if hasattr(self.backbone, 'bev_w'):
self.bev_w = self.backbone.bev_w
self.bev_h = self.backbone.bev_h
self.head = build_head(head_cfg)
def multiscale_neck(self, bev_embedding):
......@@ -97,7 +91,7 @@ class VectorMapNet(BaseMapper):
if self.last_epoch is None:
self.last_epoch = [batch, img, img_metas, valid_idx, points]
if len(valid_idx)==0:
if len(valid_idx) == 0:
batch, img, img_metas, valid_idx, points = self.last_epoch
else:
del self.last_epoch
......@@ -184,8 +178,8 @@ class VectorMapNet(BaseMapper):
return None, None, None, valid_idx, None
batch = {}
batch['det'] = format_det(polys,device)
batch['gen'] = format_gen(polys,device)
batch['det'] = format_det(polys, device)
batch['gen'] = format_gen(polys, device)
return batch, imgs, img_metas, valid_idx, points
......@@ -193,7 +187,7 @@ class VectorMapNet(BaseMapper):
pad_points = pad_sequence(points, batch_first=True)
points_mask = torch.zeros_like(pad_points[:,:,0]).bool()
points_mask = torch.zeros_like(pad_points[:, :, 0]).bool()
for i in range(len(points)):
valid_num = points[i].shape[0]
points_mask[i][:valid_num] = True
......@@ -202,15 +196,13 @@ class VectorMapNet(BaseMapper):
def format_det(polys, device):
batch = {
'class_label':[],
'batch_idx':[],
'class_label': [],
'batch_idx': [],
'bbox': [],
}
for batch_idx, poly in enumerate(polys):
keypoint_label = torch.from_numpy(poly['det_label']).to(device)
keypoint = torch.from_numpy(poly['keypoint']).to(device)
......@@ -220,8 +212,7 @@ def format_det(polys, device):
return batch
def format_gen(polys,device):
def format_gen(polys, device):
line_cls = []
polylines, polyline_masks, polyline_weights = [], [], []
bbox, line_cls, line_bs_idx = [], [], []
......@@ -230,13 +221,13 @@ def format_gen(polys,device):
# convert to cuda tensor
for k in poly.keys():
if isinstance(poly[k],np.ndarray):
if isinstance(poly[k], np.ndarray):
poly[k] = torch.from_numpy(poly[k]).to(device)
else:
poly[k] = [torch.from_numpy(v).to(device) for v in poly[k]]
line_cls += poly['gen_label']
line_bs_idx += [batch_idx]*len(poly['gen_label'])
line_bs_idx += [batch_idx] * len(poly['gen_label'])
# condition
bbox += poly['qkeypoint']
......
autonomous_driving/Online-HD-Map-Construction/src/models/transformer_utils/__init__.py
View file @ 41b18fd8
from .deformable_transformer import DeformableDetrTransformer_, DeformableDetrTransformerDecoder_
from .base_transformer import PlaceHolderEncoder
\ No newline at end of file
autonomous_driving/Online-HD-Map-Construction/src/models/transformer_utils/base_transformer.py
View file @ 41b18fd8
import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
from mmcv.cnn.bricks.registry import TRANSFORMER_LAYER_SEQUENCE
from mmcv.cnn import xavier_init, constant_init
from mmcv.cnn.bricks.registry import (ATTENTION,
TRANSFORMER_LAYER_SEQUENCE)
from mmcv.cnn.bricks.transformer import (MultiScaleDeformableAttention,
TransformerLayerSequence,
build_transformer_layer_sequence)
from mmcv.runner.base_module import BaseModule
from mmdet.models.utils.builder import TRANSFORMER
@TRANSFORMER_LAYER_SEQUENCE.register_module()
class PlaceHolderEncoder(nn.Module):
......@@ -21,5 +10,4 @@ class PlaceHolderEncoder(nn.Module):
self.embed_dims = embed_dims
def forward(self, *args, query=None, **kwargs):
return query
autonomous_driving/Online-HD-Map-Construction/src/models/transformer_utils/deformable_transformer.py
View file @ 41b18fd8
......@@ -4,18 +4,12 @@ import warnings
import torch
import torch.nn as nn
from mmcv.cnn import build_activation_layer, build_norm_layer, xavier_init
from mmcv.cnn.bricks.registry import (TRANSFORMER_LAYER,
TRANSFORMER_LAYER_SEQUENCE)
from mmcv.cnn.bricks.transformer import (BaseTransformerLayer,
TransformerLayerSequence,
build_transformer_layer_sequence)
from mmcv.runner.base_module import BaseModule
from torch.nn.init import normal_
from mmcv.cnn import xavier_init
from mmcv.cnn.bricks.registry import TRANSFORMER_LAYER_SEQUENCE
from mmcv.cnn.bricks.transformer import TransformerLayerSequence
from mmdet.models.utils.builder import TRANSFORMER
from mmdet.models.utils.transformer import Transformer
from torch.nn.init import normal_
try:
from mmcv.ops.multi_scale_deform_attn import MultiScaleDeformableAttention
......@@ -27,6 +21,7 @@ except ImportError:
from .fp16_dattn import MultiScaleDeformableAttentionFp16
def inverse_sigmoid(x, eps=1e-5):
"""Inverse function of sigmoid.
Args:
......@@ -44,6 +39,7 @@ def inverse_sigmoid(x, eps=1e-5):
x2 = (1 - x).clamp(min=eps)
return torch.log(x1 / x2)
@TRANSFORMER_LAYER_SEQUENCE.register_module()
class DeformableDetrTransformerDecoder_(TransformerLayerSequence):
"""Implements the decoder in DETR transformer.
......@@ -94,13 +90,13 @@ class DeformableDetrTransformerDecoder_(TransformerLayerSequence):
for lid, layer in enumerate(self.layers):
reference_points_input = \
reference_points[:, :, None,:self.kp_coord_dim] * \
valid_ratios[:, None,:,:self.kp_coord_dim]
reference_points[:, :, None, :self.kp_coord_dim] * \
valid_ratios[:, None, :, :self.kp_coord_dim]
# if reference_points.shape[-1] == 3 and self.kp_coord_dim==2:
output = layer(
output,
*args,
reference_points=reference_points_input[...,:self.kp_coord_dim],
reference_points=reference_points_input[..., :self.kp_coord_dim],
**kwargs)
output = output.permute(1, 0, 2)
......@@ -108,11 +104,12 @@ class DeformableDetrTransformerDecoder_(TransformerLayerSequence):
tmp = reg_branches[lid](output)
new_reference_points = tmp
new_reference_points[..., :self.kp_coord_dim] = tmp[
..., :self.kp_coord_dim] + inverse_sigmoid(reference_points)
..., :self.kp_coord_dim] + inverse_sigmoid(
reference_points)
new_reference_points = new_reference_points.sigmoid()
if reference_points.shape[-1] == 3 and self.kp_coord_dim==2:
reference_points[...,-1] = tmp[...,-1].sigmoid().detach()
reference_points[...,:self.coord_dim] = new_reference_points.detach()
if reference_points.shape[-1] == 3 and self.kp_coord_dim == 2:
reference_points[..., -1] = tmp[..., -1].sigmoid().detach()
reference_points[..., :self.coord_dim] = new_reference_points.detach()
output = output.permute(1, 0, 2)
if self.return_intermediate:
......@@ -174,7 +171,7 @@ class DeformableDetrTransformer_(Transformer):
for m in self.modules():
if isinstance(m, MultiScaleDeformableAttention):
m.init_weights()
elif isinstance(m,MultiScaleDeformableAttentionFp16):
elif isinstance(m, MultiScaleDeformableAttentionFp16):
m.init_weights()
if not self.as_two_stage:
xavier_init(self.reference_points_embed, distribution='uniform', bias=0.)
......@@ -231,7 +228,7 @@ class DeformableDetrTransformer_(Transformer):
scale = 2 * math.pi
dim_t = torch.arange(
num_pos_feats, dtype=torch.float32, device=proposals.device)
dim_t = temperature**(2 * (dim_t // 2) / num_pos_feats)
dim_t = temperature ** (2 * (dim_t // 2) / num_pos_feats)
# N, L, 4
proposals = proposals.sigmoid() * scale
# N, L, 4, 128
......@@ -317,7 +314,7 @@ class DeformableDetrTransformer_(Transformer):
spatial_shapes = torch.as_tensor(
spatial_shapes, dtype=torch.long, device=feat_flatten.device)
level_start_index = torch.cat((spatial_shapes.new_zeros(
(1, )), spatial_shapes.prod(1).cumsum(0)[:-1]))
(1,)), spatial_shapes.prod(1).cumsum(0)[:-1]))
valid_ratios = torch.stack(
[self.get_valid_ratio(m) for m in mlvl_masks], 1)
......
autonomous_driving/Online-HD-Map-Construction/src/models/transformer_utils/fp16_dattn.py
View file @ 41b18fd8
from turtle import forward
import warnings
try:
from mmcv.ops.multi_scale_deform_attn import MultiScaleDeformableAttention
except ImportError:
......@@ -7,12 +7,6 @@ except ImportError:
'`MultiScaleDeformableAttention` in MMCV has been moved to '
'`mmcv.ops.multi_scale_deform_attn`, please update your MMCV')
from mmcv.cnn.bricks.transformer import MultiScaleDeformableAttention
from mmcv.runner import force_fp32, auto_fp16
from mmcv.cnn.bricks.registry import ATTENTION
from mmcv.runner.base_module import BaseModule, ModuleList, Sequential
from mmcv.cnn.bricks.transformer import build_attention
import math
import warnings
......@@ -20,13 +14,15 @@ import warnings
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.autograd.function import Function, once_differentiable
from mmcv import deprecated_api_warning
from mmcv.cnn import constant_init, xavier_init
from mmcv.cnn.bricks.registry import ATTENTION
from mmcv.runner import BaseModule
from mmcv.cnn.bricks.transformer import build_attention
from mmcv.runner import force_fp32
from mmcv.runner.base_module import BaseModule
from mmcv.utils import ext_loader
from torch.autograd.function import Function, once_differentiable
ext_module = ext_loader.load_ext(
'_ext', ['ms_deform_attn_backward', 'ms_deform_attn_forward'])
from torch.cuda.amp import custom_bwd, custom_fwd
......@@ -35,16 +31,15 @@ from torch.cuda.amp import custom_bwd, custom_fwd
@ATTENTION.register_module()
class MultiScaleDeformableAttentionFp16(BaseModule):
def __init__(self, attn_cfg=None,init_cfg=None,**kwarg):
super(MultiScaleDeformableAttentionFp16,self).__init__(init_cfg)
def __init__(self, attn_cfg=None, init_cfg=None, **kwarg):
super(MultiScaleDeformableAttentionFp16, self).__init__(init_cfg)
# import ipdb; ipdb.set_trace()
self.deformable_attention = build_attention(attn_cfg)
self.deformable_attention.init_weights()
self.fp16_enabled = False
@force_fp32(apply_to=('query', 'key', 'value', 'query_pos', 'reference_points','identity'))
@force_fp32(apply_to=('query', 'key', 'value', 'query_pos', 'reference_points', 'identity'))
def forward(self, query,
key=None,
value=None,
......@@ -64,8 +59,7 @@ class MultiScaleDeformableAttentionFp16(BaseModule):
key_padding_mask=key_padding_mask,
reference_points=reference_points,
spatial_shapes=spatial_shapes,
level_start_index=level_start_index,**kwargs)
level_start_index=level_start_index, **kwargs)
class MultiScaleDeformableAttnFunctionFp32(Function):
......@@ -118,7 +112,7 @@ class MultiScaleDeformableAttnFunctionFp32(Function):
Tuple[Tensor]: Gradient
of input tensors in forward.
"""
value, value_spatial_shapes, value_level_start_index,\
value, value_spatial_shapes, value_level_start_index, \
sampling_locations, attention_weights = ctx.saved_tensors
grad_value = torch.zeros_like(value)
grad_sampling_loc = torch.zeros_like(sampling_locations)
......@@ -161,7 +155,7 @@ def multi_scale_deformable_attn_pytorch(value, value_spatial_shapes,
"""
bs, _, num_heads, embed_dims = value.shape
_, num_queries, num_heads, num_levels, num_points, _ =\
_, num_queries, num_heads, num_levels, num_points, _ = \
sampling_locations.shape
value_list = value.split([H_ * W_ for H_, W_ in value_spatial_shapes],
dim=1)
......
autonomous_driving/Online-HD-Map-Construction/tools/evaluate_submission.py
View file @ 41b18fd8
import sys
import os
import sys
sys.path.append(os.path.abspath('.'))
from src.datasets.evaluation.vector_eval import VectorEvaluate
import argparse
from src.datasets.evaluation.vector_eval import VectorEvaluate
def parse_args():
parser = argparse.ArgumentParser(
description='Evaluate a submission file')
......@@ -17,12 +20,14 @@ def parse_args():
args = parser.parse_args()
return args
def main(args):
evaluator = VectorEvaluate(args.gt, n_workers=0)
results = evaluator.evaluate(args.submission)
print(results)
if __name__ == '__main__':
args = parse_args()
main(args)
autonomous_driving/Online-HD-Map-Construction/tools/mmdet_test.py
View file @ 41b18fd8
......@@ -9,7 +9,6 @@ import torch
import torch.distributed as dist
from mmcv.image import tensor2imgs
from mmcv.runner import get_dist_info
from mmdet.core import encode_mask_results
......@@ -120,7 +119,7 @@ def collect_results_cpu(result_part, size, tmpdir=None):
if tmpdir is None:
MAX_LEN = 512
# 32 is whitespace
dir_tensor = torch.full((MAX_LEN, ),
dir_tensor = torch.full((MAX_LEN,),
32,
dtype=torch.uint8,
device='cuda')
......
autonomous_driving/Online-HD-Map-Construction/tools/mmdet_train.py
View file @ 41b18fd8
......@@ -8,7 +8,6 @@ from mmcv.runner import (HOOKS, DistSamplerSeedHook, EpochBasedRunner,
Fp16OptimizerHook, OptimizerHook, build_optimizer,
build_runner)
from mmcv.utils import build_from_cfg
from mmdet.core import DistEvalHook, EvalHook
from mmdet.datasets import (build_dataloader, build_dataset,
replace_ImageToTensor)
......
autonomous_driving/Online-HD-Map-Construction/tools/test.py
View file @ 41b18fd8
import argparse
import mmcv
import os
import os.path as osp
import torch
import warnings
from mmcv import Config, DictAction
from mmcv import Config
from mmcv.cnn import fuse_conv_bn
from mmcv.parallel import MMDataParallel, MMDistributedDataParallel
from mmcv.runner import (get_dist_info, init_dist, load_checkpoint,
wrap_fp16_model)
from mmdet3d.apis import single_gpu_test
from mmdet3d.datasets import build_dataloader, build_dataset
from mmdet3d.models import build_model
from mmdet.datasets import replace_ImageToTensor
from mmdet_test import multi_gpu_test
from mmdet_train import set_random_seed
from mmdet.datasets import replace_ImageToTensor
def parse_args():
......@@ -106,8 +104,8 @@ def main():
plg_lib = importlib.import_module(_module_path)
plugin_dirs = cfg.plugin_dir
if not isinstance(plugin_dirs,list):
plugin_dirs = [plugin_dirs,]
if not isinstance(plugin_dirs, list):
plugin_dirs = [plugin_dirs, ]
for plugin_dir in plugin_dirs:
import_path(plugin_dir)
......@@ -154,7 +152,7 @@ def main():
osp.splitext(osp.basename(args.config))[0])
cfg_data_dict.work_dir = cfg.work_dir
print('work_dir: ',cfg.work_dir)
print('work_dir: ', cfg.work_dir)
dataset = build_dataset(cfg_data_dict)
data_loader = build_dataloader(
dataset,
......
autonomous_driving/Online-HD-Map-Construction/tools/train.py
View file @ 41b18fd8
......@@ -2,26 +2,25 @@ from __future__ import division
import argparse
import copy
import mmcv
import os
import time
import torch
import warnings
from mmcv import Config, DictAction
from mmcv.runner import get_dist_info, init_dist
from os import path as osp
import mmcv
import torch
from mmcv import Config, DictAction
from mmcv.runner import get_dist_info, init_dist
from mmdet import __version__ as mmdet_version
from mmdet3d import __version__ as mmdet3d_version
from mmdet3d.apis import train_model
from mmdet3d.datasets import build_dataset
# from builder import build_model
from mmdet3d.models import build_model
from mmdet3d.utils import collect_env, get_root_logger
from mmseg import __version__ as mmseg_version
# warper
from mmdet_train import set_random_seed
# from builder import build_model
from mmdet3d.models import build_model
from mmseg import __version__ as mmseg_version
def parse_args():
......@@ -126,8 +125,8 @@ def main():
plg_lib = importlib.import_module(_module_path)
plugin_dirs = cfg.plugin_dir
if not isinstance(plugin_dirs,list):
plugin_dirs = [plugin_dirs,]
if not isinstance(plugin_dirs, list):
plugin_dirs = [plugin_dirs, ]
for plugin_dir in plugin_dirs:
import_path(plugin_dir)
......
autonomous_driving/Online-HD-Map-Construction/tools/visualization/renderer.py
View file @ 41b18fd8
import os.path as osp
import os
import numpy as np
import copy
import os
import os.path as osp
import cv2
import matplotlib.pyplot as plt
import numpy as np
from PIL import Image
from shapely.geometry import LineString
def remove_nan_values(uv):
is_u_valid = np.logical_not(np.isnan(uv[:, 0]))
is_v_valid = np.logical_not(np.isnan(uv[:, 1]))
......@@ -15,6 +17,7 @@ def remove_nan_values(uv):
uv_valid = uv[is_uv_valid]
return uv_valid
def points_ego2img(pts_ego, extrinsics, intrinsics):
pts_ego_4d = np.concatenate([pts_ego, np.ones([len(pts_ego), 1])], axis=-1)
pts_cam_4d = extrinsics @ pts_ego_4d.T
......@@ -26,6 +29,7 @@ def points_ego2img(pts_ego, extrinsics, intrinsics):
return uv, depth
def interp_fixed_dist(line, sample_dist):
''' Interpolate a line at fixed interval.
......@@ -39,13 +43,14 @@ def interp_fixed_dist(line, sample_dist):
distances = list(np.arange(sample_dist, line.length, sample_dist))
# make sure to sample at least two points when sample_dist > line.length
distances = [0,] + distances + [line.length,]
distances = [0, ] + distances + [line.length, ]
sampled_points = np.array([list(line.interpolate(distance).coords)
for distance in distances]).squeeze()
return sampled_points
def draw_polyline_ego_on_img(polyline_ego, img_bgr, extrinsics, intrinsics, color_bgr, thickness):
# if 2-dimension, assume z=0
if polyline_ego.shape[1] == 2:
......@@ -103,6 +108,7 @@ def draw_polyline_ego_on_img(polyline_ego, img_bgr, extrinsics, intrinsics, colo
# thickness_px=thickness,
# )
def draw_visible_polyline_cv2(line, valid_pts_bool, image, color, thickness_px):
"""Draw a polyline onto an image using given line segments.
Args:
......@@ -147,8 +153,9 @@ COLOR_MAPS_PLT = {
}
CAM_NAMES_AV2 = ['ring_front_center', 'ring_front_right', 'ring_front_left',
'ring_rear_right','ring_rear_left', 'ring_side_right', 'ring_side_left',
]
'ring_rear_right', 'ring_rear_left', 'ring_side_right', 'ring_side_left',
]
class Renderer(object):
"""Render map elements on image views.
......@@ -171,8 +178,8 @@ class Renderer(object):
map_path = os.path.join(out_dir, 'map.jpg')
plt.figure(figsize=(self.roi_size[0], self.roi_size[1]))
plt.xlim(-self.roi_size[0]/2 - 1, self.roi_size[0]/2 + 1)
plt.ylim(-self.roi_size[1]/2 - 1, self.roi_size[1]/2 + 1)
plt.xlim(-self.roi_size[0] / 2 - 1, self.roi_size[0] / 2 + 1)
plt.ylim(-self.roi_size[1] / 2 - 1, self.roi_size[1] / 2 + 1)
plt.axis('off')
plt.imshow(car_img, extent=[-1.5, 1.5, -1.2, 1.2])
......
autonomous_driving/Online-HD-Map-Construction/tools/visualization/visualize.py
View file @ 41b18fd8
import argparse
import mmcv
from mmcv import Config
import os
import mmcv
from renderer import Renderer
CAT2ID = {
......@@ -14,6 +14,7 @@ ID2CAT = {v: k for k, v in CAT2ID.items()}
ROI_SIZE = (60, 30)
def parse_args():
parser = argparse.ArgumentParser(
description='Visualize groundtruth and results')
......@@ -37,6 +38,7 @@ def parse_args():
return args
def import_plugin(cfg):
'''
import modules, registry will be update
......@@ -59,8 +61,8 @@ def import_plugin(cfg):
plg_lib = importlib.import_module(_module_path)
plugin_dirs = cfg.plugin_dir
if not isinstance(plugin_dirs,list):
plugin_dirs = [plugin_dirs,]
if not isinstance(plugin_dirs, list):
plugin_dirs = [plugin_dirs, ]
for plugin_dir in plugin_dirs:
import_path(plugin_dir)
......@@ -74,6 +76,7 @@ def import_plugin(cfg):
print(f'importing {_module_path}/')
plg_lib = importlib.import_module(_module_path)
def main(args):
log_id = args.log_id
ann = mmcv.load(args.ann_file)
......
autonomous_driving/README.md
View file @ 41b18fd8
<div id="top" align="center">
# InternImage for CVPR 2023 Workshop on End-to-End Autonomous Driving
</div>
</div>
## 1. InternImage-based Baseline for CVPR23 Occupancy Prediction Challenge
We achieve an improvement of 1.44 in MIOU baseline by leveraging the InterImage-based model.
model name|weight| mIoU | others | barrier | bicycle | bus | car | construction_vehicle | motorcycle | pedestrian | traffic_cone | trailer | truck | driveable_surface | other_flat | sidewalk | terrain | manmade | vegetation |
----|:----------:| :--: | :--: | :--: | :--: | :--: | :--: | :--: | :--: | :--: | :--: | :--: | :--: | :--: | :--: | :----------------------: | :---: | :------: | :------: |
bevformer_intern-s_occ|[Google Drive](https://drive.google.com/file/d/1LV9K8hrskKf51xY1wbqTKzK7WZmVXEV_/view?usp=sharing)| 25.11 | 6.93 | 35.57 | 10.40 | 35.97 | 41.23 | 13.72 | 20.30 | 21.10 | 18.34 | 19.18 | 28.64 | 49.82 | 30.74 | 31.00 | 27.44 | 19.29 | 17.29 |
bevformer_base_occ|[Google Drive](https://drive.google.com/file/d/1NyoiosafAmne1qiABeNOPXR-P-y0i7_I/view?usp=share_link)| 23.67 | 5.03 | 38.79 | 9.98 | 34.41 | 41.09 | 13.24 | 16.50 | 18.15 | 17.83 | 18.66 | 27.70 | 48.95 | 27.73 | 29.08 | 25.38 | 15.41 | 14.46 |
| model name | weight | mIoU | others | barrier | bicycle | bus | car | construction_vehicle | motorcycle | pedestrian | traffic_cone | trailer | truck | driveable_surface | other_flat | sidewalk | terrain | manmade | vegetation |
| ---------------------- | :---------------------------------------------------------------------------------------------------: | :---: | :----: | :-----: | :-----: | :---: | :---: | :------------------: | :--------: | :--------: | :----------: | :-----: | :---: | :---------------: | :--------: | :------: | :-----: | :-----: | :--------: |
| bevformer_intern-s_occ | [Google Drive](https://drive.google.com/file/d/1LV9K8hrskKf51xY1wbqTKzK7WZmVXEV_/view?usp=sharing) | 25.11 | 6.93 | 35.57 | 10.40 | 35.97 | 41.23 | 13.72 | 20.30 | 21.10 | 18.34 | 19.18 | 28.64 | 49.82 | 30.74 | 31.00 | 27.44 | 19.29 | 17.29 |
| bevformer_base_occ | [Google Drive](https://drive.google.com/file/d/1NyoiosafAmne1qiABeNOPXR-P-y0i7_I/view?usp=share_link) | 23.67 | 5.03 | 38.79 | 9.98 | 34.41 | 41.09 | 13.24 | 16.50 | 18.15 | 17.83 | 18.66 | 27.70 | 48.95 | 27.73 | 29.08 | 25.38 | 15.41 | 14.46 |
### Get Started
please refer to [README.md](./occupancy_prediction/README.md)
please refer to [README.md](./occupancy_prediction/README.md)
## 2. InternImage-based Baseline for Online HD Map Construction Challenge For Autonomous Driving
By incorporating the InterImage-based model, we observe an enhancement of 6.56 in mAP baseline.
model name|weight|$\mathrm{mAP}$ | $\mathrm{AP}_{pc}$ | $\mathrm{AP}_{div}$ | $\mathrm{AP}_{bound}$ |
----|:----------:| :--: | :--: | :--: | :--: |
vectormapnet_intern|[Checkpoint](https://github.com/OpenGVLab/InternImage/releases/download/track_model/vectormapnet_internimage.pth) | 49.35 | 45.05 | 56.78 | 46.22 |
vectormapnet_base|[Google Drive](https://drive.google.com/file/d/16D1CMinwA8PG1sd9PV9_WtHzcBohvO-D/view) | 42.79 | 37.22 | 50.47 | 40.68 |
| model name | weight | $\\mathrm{mAP}$ | $\\mathrm{AP}\_{pc}$ | $\\mathrm{AP}\_{div}$ | $\\mathrm{AP}\_{bound}$ |
| ------------------- | :---------------------------------------------------------------------------------------------------------------: | :-------------: | :------------------: | :-------------------: | :---------------------: |
| vectormapnet_intern | [Checkpoint](https://github.com/OpenGVLab/InternImage/releases/download/track_model/vectormapnet_internimage.pth) | 49.35 | 45.05 | 56.78 | 46.22 |
| vectormapnet_base | [Google Drive](https://drive.google.com/file/d/16D1CMinwA8PG1sd9PV9_WtHzcBohvO-D/view) | 42.79 | 37.22 | 50.47 | 40.68 |
### Get Started
please refer to [README.md](Online-HD-Map-Construction/README.md)
please refer to [README.md](Online-HD-Map-Construction/README.md)
## 3. InternImage-based Baseline for CVPR23 OpenLane-V2 Challenge
Through the implementation of the InterImage-based model, we achieve an advancement of 0.009 in F-score baseline.
Through the implementation of the InterImage-based model, we achieve an advancement of 0.009 in F-score baseline.
| | OpenLane-V2 Score | DET<sub>l</sub> | DET<sub>t</sub> | TOP<sub>ll</sub> | TOP<sub>lt</sub> | F-Score |
|-------------|-------------------|-----------------|-----------------|------------------|------------------|---------|
| ----------- | ----------------- | --------------- | --------------- | ---------------- | ---------------- | ------- |
| base r50 | 0.292 | 0.183 | 0.457 | 0.022 | 0.143 | 0.215 |
| InternImage | 0.325 | 0.194 | 0.537 | 0.02 | 0.17 | 0.224 |
### Get Started
please refer to [README.md](./openlane-v2/README.md)
please refer to [README.md](./openlane-v2/README.md)
autonomous_driving/occupancy_prediction/CODE_OF_CONDUCT.md
View file @ 41b18fd8
......@@ -2,127 +2,101 @@
## Our Pledge
We as members, contributors, and leaders pledge to make participation in our
community a harassment-free experience for everyone, regardless of age, body
size, visible or invisible disability, ethnicity, sex characteristics, gender
identity and expression, level of experience, education, socio-economic status,
nationality, personal appearance, race, religion, or sexual identity
and orientation.
We as members, contributors, and leaders pledge to make participation in our community a harassment-free experience for
everyone, regardless of age, body size, visible or invisible disability, ethnicity, sex characteristics, gender identity
and expression, level of experience, education, socio-economic status, nationality, personal appearance, race, religion,
or sexual identity and orientation.
We pledge to act and interact in ways that contribute to an open, welcoming,
diverse, inclusive, and healthy community.
We pledge to act and interact in ways that contribute to an open, welcoming, diverse, inclusive, and healthy community.
## Our Standards
Examples of behavior that contributes to a positive environment for our
community include:
Examples of behavior that contributes to a positive environment for our community include:
* Demonstrating empathy and kindness toward other people
* Being respectful of differing opinions, viewpoints, and experiences
* Giving and gracefully accepting constructive feedback
* Accepting responsibility and apologizing to those affected by our mistakes,
and learning from the experience
* Focusing on what is best not just for us as individuals, but for the
overall community
- Demonstrating empathy and kindness toward other people
- Being respectful of differing opinions, viewpoints, and experiences
- Giving and gracefully accepting constructive feedback
- Accepting responsibility and apologizing to those affected by our mistakes, and learning from the experience
- Focusing on what is best not just for us as individuals, but for the overall community
Examples of unacceptable behavior include:
* The use of sexualized language or imagery, and sexual attention or
advances of any kind
* Trolling, insulting or derogatory comments, and personal or political attacks
* Public or private harassment
* Publishing others' private information, such as a physical or email
address, without their explicit permission
* Other conduct which could reasonably be considered inappropriate in a
professional setting
- The use of sexualized language or imagery, and sexual attention or advances of any kind
- Trolling, insulting or derogatory comments, and personal or political attacks
- Public or private harassment
- Publishing others' private information, such as a physical or email address, without their explicit permission
- Other conduct which could reasonably be considered inappropriate in a professional setting
## Enforcement Responsibilities
Community leaders are responsible for clarifying and enforcing our standards of
acceptable behavior and will take appropriate and fair corrective action in
response to any behavior that they deem inappropriate, threatening, offensive,
Community leaders are responsible for clarifying and enforcing our standards of acceptable behavior and will take
appropriate and fair corrective action in response to any behavior that they deem inappropriate, threatening, offensive,
or harmful.
Community leaders have the right and responsibility to remove, edit, or reject
comments, commits, code, wiki edits, issues, and other contributions that are
not aligned to this Code of Conduct, and will communicate reasons for moderation
decisions when appropriate.
Community leaders have the right and responsibility to remove, edit, or reject comments, commits, code, wiki edits,
issues, and other contributions that are not aligned to this Code of Conduct, and will communicate reasons for
moderation decisions when appropriate.
## Scope
This Code of Conduct applies within all community spaces, and also applies when
an individual is officially representing the community in public spaces.
Examples of representing our community include using an official e-mail address,
posting via an official social media account, or acting as an appointed
representative at an online or offline event.
This Code of Conduct applies within all community spaces, and also applies when an individual is officially representing
the community in public spaces. Examples of representing our community include using an official e-mail address, posting
via an official social media account, or acting as an appointed representative at an online or offline event.
## Enforcement
Instances of abusive, harassing, or otherwise unacceptable behavior may be
reported to the community leaders responsible for enforcement at
contact@opendrivelab.com.
All complaints will be reviewed and investigated promptly and fairly.
Instances of abusive, harassing, or otherwise unacceptable behavior may be reported to the community leaders responsible
for enforcement at contact@opendrivelab.com. All complaints will be reviewed and investigated promptly and fairly.
All community leaders are obligated to respect the privacy and security of the
reporter of any incident.
All community leaders are obligated to respect the privacy and security of the reporter of any incident.
## Enforcement Guidelines
Community leaders will follow these Community Impact Guidelines in determining
the consequences for any action they deem in violation of this Code of Conduct:
Community leaders will follow these Community Impact Guidelines in determining the consequences for any action they deem
in violation of this Code of Conduct:
### 1. Correction
**Community Impact**: Use of inappropriate language or other behavior deemed
unprofessional or unwelcome in the community.
**Community Impact**: Use of inappropriate language or other behavior deemed unprofessional or unwelcome in the
community.
**Consequence**: A private, written warning from community leaders, providing
clarity around the nature of the violation and an explanation of why the
behavior was inappropriate. A public apology may be requested.
**Consequence**: A private, written warning from community leaders, providing clarity around the nature of the violation
and an explanation of why the behavior was inappropriate. A public apology may be requested.
### 2. Warning
**Community Impact**: A violation through a single incident or series
of actions.
**Community Impact**: A violation through a single incident or series of actions.
**Consequence**: A warning with consequences for continued behavior. No
interaction with the people involved, including unsolicited interaction with
those enforcing the Code of Conduct, for a specified period of time. This
includes avoiding interactions in community spaces as well as external channels
like social media. Violating these terms may lead to a temporary or
permanent ban.
**Consequence**: A warning with consequences for continued behavior. No interaction with the people involved, including
unsolicited interaction with those enforcing the Code of Conduct, for a specified period of time. This includes avoiding
interactions in community spaces as well as external channels like social media. Violating these terms may lead to a
temporary or permanent ban.
### 3. Temporary Ban
**Community Impact**: A serious violation of community standards, including
sustained inappropriate behavior.
**Community Impact**: A serious violation of community standards, including sustained inappropriate behavior.
**Consequence**: A temporary ban from any sort of interaction or public
communication with the community for a specified period of time. No public or
private interaction with the people involved, including unsolicited interaction
with those enforcing the Code of Conduct, is allowed during this period.
Violating these terms may lead to a permanent ban.
**Consequence**: A temporary ban from any sort of interaction or public communication with the community for a specified
period of time. No public or private interaction with the people involved, including unsolicited interaction with those
enforcing the Code of Conduct, is allowed during this period. Violating these terms may lead to a permanent ban.
### 4. Permanent Ban
**Community Impact**: Demonstrating a pattern of violation of community
standards, including sustained inappropriate behavior, harassment of an
individual, or aggression toward or disparagement of classes of individuals.
**Community Impact**: Demonstrating a pattern of violation of community standards, including sustained inappropriate
behavior, harassment of an individual, or aggression toward or disparagement of classes of individuals.
**Consequence**: A permanent ban from any sort of public interaction within
the community.
**Consequence**: A permanent ban from any sort of public interaction within the community.
## Attribution
This Code of Conduct is adapted from the [Contributor Covenant][homepage],
version 2.0, available at
This Code of Conduct is adapted from the [Contributor Covenant][homepage], version 2.0, available at
https://www.contributor-covenant.org/version/2/0/code_of_conduct.html.
Community Impact Guidelines were inspired by [Mozilla's code of conduct
enforcement ladder](https://github.com/mozilla/diversity).
[homepage]: https://www.contributor-covenant.org
Community Impact Guidelines were inspired
by [Mozilla's code of conduct enforcement ladder](https://github.com/mozilla/diversity).
For answers to common questions about this code of conduct, see the FAQ at
https://www.contributor-covenant.org/faq. Translations are available at
https://www.contributor-covenant.org/translations.
[homepage]: https://www.contributor-covenant.org
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment