Refactor optimizer and samplers

.gitlab-ci.yml

@@ -25,6 +25,7 @@ before_script:
   stage: test
   script:
     - echo "Start building..."
+    - pip install "git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI"
     - pip install git+https://github.com/open-mmlab/mmdetection.git@v2.0
     - python -c "import mmdet; print(mmdet.__version__)"
     - pip install -v -e .[all]

configs/kitti/dv_mvx-v2_second_secfpn_fpn-fusion_adamw_2x8_80e_kitti-3d-3class.py

@@ -83,7 +83,7 @@ model = dict(
         anchor_rotations=[0, 1.57],
         diff_rad_by_sin=True,
         assign_per_class=True,
-        bbox_coder=dict(type='ResidualCoder', ),
+        bbox_coder=dict(type='Residual3DBoxCoder', ),
         loss_cls=dict(
             type='FocalLoss',
             use_sigmoid=True,

configs/kitti/dv_pointpillars_secfpn_6x8_160e_kitti-3d-car.py

@@ -48,7 +48,7 @@ model = dict(
         anchor_sizes=[[1.6, 3.9, 1.56]],
         anchor_rotations=[0, 1.57],
         diff_rad_by_sin=True,
-        bbox_coder=dict(type='ResidualCoder', ),
+        bbox_coder=dict(type='Residual3DBoxCoder', ),
         loss_cls=dict(
             type='FocalLoss',
             use_sigmoid=True,

configs/kitti/dv_second_secfpn_2x8_cosine_80e_kitti-3d-3class.py

@@ -52,7 +52,7 @@ model = dict(
         diff_rad_by_sin=True,
         assigner_per_size=True,
         assign_per_class=True,
-        bbox_coder=dict(type='ResidualCoder', ),
+        bbox_coder=dict(type='Residual3DBoxCoder', ),
         loss_cls=dict(
             type='FocalLoss',
             use_sigmoid=True,

configs/kitti/dv_second_secfpn_6x8_80e_kitti-3d-car.py

@@ -46,7 +46,7 @@ model = dict(
         anchor_sizes=[[1.6, 3.9, 1.56]],
         anchor_rotations=[0, 1.57],
         diff_rad_by_sin=True,
-        bbox_coder=dict(type='ResidualCoder', ),
+        bbox_coder=dict(type='Residual3DBoxCoder', ),
         loss_cls=dict(
             type='FocalLoss',
             use_sigmoid=True,

configs/kitti/faster_rcnn_r50_fpn_caffe_1x_kitti-2d-3class_coco-3x-pretrain.py

@@ -29,7 +29,7 @@ model = dict(
         target_stds=[1.0, 1.0, 1.0, 1.0],
         loss_cls=dict(
             type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
-        loss_bbox=dict(type='SmoothL1Loss', beta=0.0, loss_weight=1.0)),
+        loss_bbox=dict(type='L1Loss', loss_weight=1.0)),
     roi_head=dict(
         type='StandardRoIHead',
         bbox_roi_extractor=dict(

configs/kitti/hv_pointpillars_secfpn_6x8_160e_kitti-3d-car.py

@@ -47,7 +47,7 @@ model = dict(
         anchor_sizes=[[1.6, 3.9, 1.56]],
         anchor_rotations=[0, 1.57],
         diff_rad_by_sin=True,
-        bbox_coder=dict(type='ResidualCoder', ),
+        bbox_coder=dict(type='Residual3DBoxCoder', ),
         loss_cls=dict(
             type='FocalLoss',
             use_sigmoid=True,

configs/kitti/hv_second_secfpn_6x8_80e_kitti-3d-car.py

@@ -46,7 +46,7 @@ model = dict(
         anchor_sizes=[[1.6, 3.9, 1.56]],
         anchor_rotations=[0, 1.57],
         diff_rad_by_sin=True,
-        bbox_coder=dict(type='ResidualCoder', ),
+        bbox_coder=dict(type='Residual3DBoxCoder', ),
         loss_cls=dict(
             type='FocalLoss',
             use_sigmoid=True,

configs/nus/faster_rcnn_r50_fpn_caffe_2x8_1x_nus.py

@@ -29,7 +29,7 @@ model = dict(
         target_stds=[1.0, 1.0, 1.0, 1.0],
         loss_cls=dict(
             type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
-        loss_bbox=dict(type='SmoothL1Loss', beta=0.0, loss_weight=1.0)),
+        loss_bbox=dict(type='L1Loss', loss_weight=1.0)),
     roi_head=dict(
         type='StandardRoIHead',
         bbox_roi_extractor=dict(

configs/nus/hv_pointpillars_secfpn_sbn-all_4x8_20e_nus-3d.py

@@ -77,7 +77,7 @@ model = dict(
         diff_rad_by_sin=True,
         dir_offset=0.7854,  # pi/4
         dir_limit_offset=0,
-        bbox_coder=dict(type='ResidualCoder', ),
+        bbox_coder=dict(type='Residual3DBoxCoder', ),
         loss_cls=dict(
             type='FocalLoss',
             use_sigmoid=True,

configs/nus/retinanet_r50_fpn_caffe_2x8_1x_nus.py

@@ -38,7 +38,7 @@ model = dict(
             gamma=2.0,
             alpha=0.25,
             loss_weight=1.0),
-        loss_bbox=dict(type='SmoothL1Loss', beta=0.0, loss_weight=1.0)))
+        loss_bbox=dict(type='L1Loss', loss_weight=1.0)))
 # training and testing settings
 train_cfg = dict(
     assigner=dict(

mmdet3d/apis/train.py

-import torch
-from mmcv.parallel import MMDataParallel, MMDistributedDataParallel
-from mmcv.runner import DistSamplerSeedHook, Runner
-
-from mmdet3d.core import build_optimizer
-from mmdet3d.datasets import build_dataloader, build_dataset
 from mmdet.apis.train import parse_losses
-from mmdet.core import (DistEvalHook, DistOptimizerHook, EvalHook,
-                        Fp16OptimizerHook)
-from mmdet.utils import get_root_logger
 
 
 def batch_processor(model, data, train_mode):
@@ -36,164 +27,3 @@ def batch_processor(model, data, train_mode):
     outputs = dict(loss=loss, log_vars=log_vars, num_samples=num_samples)
 
     return outputs
-
-
-def train_detector(model,
-                   dataset,
-                   cfg,
-                   distributed=False,
-                   validate=False,
-                   timestamp=None,
-                   meta=None):
-    logger = get_root_logger(cfg.log_level)
-
-    # start training
-    if distributed:
-        _dist_train(
-            model,
-            dataset,
-            cfg,
-            validate=validate,
-            logger=logger,
-            timestamp=timestamp,
-            meta=meta)
-    else:
-        _non_dist_train(
-            model,
-            dataset,
-            cfg,
-            validate=validate,
-            logger=logger,
-            timestamp=timestamp,
-            meta=meta)
-
-
-def _dist_train(model,
-                dataset,
-                cfg,
-                validate=False,
-                logger=None,
-                timestamp=None,
-                meta=None):
-    # prepare data loaders
-    dataset = dataset if isinstance(dataset, (list, tuple)) else [dataset]
-    data_loaders = [
-        build_dataloader(
-            ds,
-            cfg.data.samples_per_gpu,
-            cfg.data.workers_per_gpu,
-            dist=True,
-            seed=cfg.seed) for ds in dataset
-    ]
-    # put model on gpus
-    find_unused_parameters = cfg.get('find_unused_parameters', False)
-    # Sets the `find_unused_parameters` parameter in
-    # torch.nn.parallel.DistributedDataParallel
-    model = MMDistributedDataParallel(
-        model.cuda(),
-        device_ids=[torch.cuda.current_device()],
-        broadcast_buffers=False,
-        find_unused_parameters=find_unused_parameters)
-
-    # build runner
-    optimizer = build_optimizer(model, cfg.optimizer)
-    runner = Runner(
-        model,
-        batch_processor,
-        optimizer,
-        cfg.work_dir,
-        logger=logger,
-        meta=meta)
-    # an ugly walkaround to make the .log and .log.json filenames the same
-    runner.timestamp = timestamp
-
-    # fp16 setting
-    fp16_cfg = cfg.get('fp16', None)
-    if fp16_cfg is not None:
-        optimizer_config = Fp16OptimizerHook(**cfg.optimizer_config,
-                                             **fp16_cfg)
-    else:
-        optimizer_config = DistOptimizerHook(**cfg.optimizer_config)
-
-    # register hooks
-    runner.register_training_hooks(cfg.lr_config, optimizer_config,
-                                   cfg.checkpoint_config, cfg.log_config)
-    runner.register_hook(DistSamplerSeedHook())
-    # register eval hooks
-    if validate:
-        val_dataset = build_dataset(cfg.data.val, dict(test_mode=True))
-        val_dataloader = build_dataloader(
-            val_dataset,
-            samples_per_gpu=1,
-            workers_per_gpu=cfg.data.workers_per_gpu,
-            dist=True,
-            shuffle=False)
-        eval_cfg = cfg.get('evaluation', {})
-        runner.register_hook(DistEvalHook(val_dataloader, **eval_cfg))
-
-    if cfg.resume_from:
-        runner.resume(cfg.resume_from)
-    elif cfg.load_from:
-        runner.load_checkpoint(cfg.load_from)
-    runner.run(data_loaders, cfg.workflow, cfg.total_epochs)
-
-
-def _non_dist_train(model,
-                    dataset,
-                    cfg,
-                    validate=False,
-                    logger=None,
-                    timestamp=None,
-                    meta=None):
-    # prepare data loaders
-    dataset = dataset if isinstance(dataset, (list, tuple)) else [dataset]
-    data_loaders = [
-        build_dataloader(
-            ds,
-            cfg.data.samples_per_gpu,
-            cfg.data.workers_per_gpu,
-            cfg.gpus,
-            dist=False,
-            seed=cfg.seed) for ds in dataset
-    ]
-    # put model on gpus
-    model = MMDataParallel(model, device_ids=range(cfg.gpus)).cuda()
-
-    # build runner
-    optimizer = build_optimizer(model, cfg.optimizer)
-    runner = Runner(
-        model,
-        batch_processor,
-        optimizer,
-        cfg.work_dir,
-        logger=logger,
-        meta=meta)
-    # an ugly walkaround to make the .log and .log.json filenames the same
-    runner.timestamp = timestamp
-    # fp16 setting
-    fp16_cfg = cfg.get('fp16', None)
-    if fp16_cfg is not None:
-        optimizer_config = Fp16OptimizerHook(
-            **cfg.optimizer_config, **fp16_cfg, distributed=False)
-    else:
-        optimizer_config = cfg.optimizer_config
-    runner.register_training_hooks(cfg.lr_config, optimizer_config,
-                                   cfg.checkpoint_config, cfg.log_config)
-
-    # register eval hooks
-    if validate:
-        val_dataset = build_dataset(cfg.data.val, dict(test_mode=True))
-        val_dataloader = build_dataloader(
-            val_dataset,
-            samples_per_gpu=1,
-            workers_per_gpu=cfg.data.workers_per_gpu,
-            dist=False,
-            shuffle=False)
-        eval_cfg = cfg.get('evaluation', {})
-        runner.register_hook(EvalHook(val_dataloader, **eval_cfg))
-
-    if cfg.resume_from:
-        runner.resume(cfg.resume_from)
-    elif cfg.load_from:
-        runner.load_checkpoint(cfg.load_from)
-    runner.run(data_loaders, cfg.workflow, cfg.total_epochs)

mmdet3d/core/anchor/__init__.py

-from .anchor_generator import (AlignedAnchorGeneratorRange, AnchorGenerator,
+from .anchor_3d_generator import (AlignedAnchorGeneratorRange,
                                   AnchorGeneratorRange)
 
 __all__ = [
-    'AnchorGenerator', 'AlignedAnchorGeneratorRange', 'AnchorGeneratorRange',
+    'AlignedAnchorGeneratorRange', 'AnchorGeneratorRange',
     'build_anchor_generator'
 ]
 
 
 def build_anchor_generator(cfg, **kwargs):
-    from . import anchor_generator
+    from . import anchor_3d_generator
     import mmcv
     if isinstance(cfg, dict):
         return mmcv.runner.obj_from_dict(
-            cfg, anchor_generator, default_args=kwargs)
+            cfg, anchor_3d_generator, default_args=kwargs)
     else:
         raise TypeError('Invalid type {} for building a sampler'.format(
             type(cfg)))

mmdet3d/core/anchor/anchor_generator.py → mmdet3d/core/anchor/anchor_3d_generator.py

 import torch
 
 
-class AnchorGenerator(object):
-    """
-    Examples:
-        >>> from mmdet.core import AnchorGenerator
-        >>> self = AnchorGenerator(9, [1.], [1.])
-        >>> all_anchors = self.grid_anchors((2, 2), device='cpu')
-        >>> print(all_anchors)
-        tensor([[ 0.,  0.,  8.,  8.],
-                [16.,  0., 24.,  8.],
-                [ 0., 16.,  8., 24.],
-                [16., 16., 24., 24.]])
-    """
-
-    def __init__(self, base_size, scales, ratios, scale_major=True, ctr=None):
-        self.base_size = base_size
-        self.scales = torch.Tensor(scales)
-        self.ratios = torch.Tensor(ratios)
-        self.scale_major = scale_major
-        self.ctr = ctr
-        self.base_anchors = self.gen_base_anchors()
-
-    @property
-    def num_base_anchors(self):
-        return self.base_anchors.size(0)
-
-    def gen_base_anchors(self):
-        w = self.base_size
-        h = self.base_size
-
-        h_ratios = torch.sqrt(self.ratios)
-        w_ratios = 1 / h_ratios
-        if self.scale_major:
-            ws = (w * w_ratios[:, None] * self.scales[None, :]).view(-1)
-            hs = (h * h_ratios[:, None] * self.scales[None, :]).view(-1)
-        else:
-            ws = (w * self.scales[:, None] * w_ratios[None, :]).view(-1)
-            hs = (h * self.scales[:, None] * h_ratios[None, :]).view(-1)
-
-        # yapf: disable
-        base_anchors = torch.stack(
-            [
-                -0.5 * ws, -0.5 * hs,
-                0.5 * ws, 0.5 * hs
-            ],
-            dim=-1)
-        # yapf: enable
-        return base_anchors
-
-    def _meshgrid(self, x, y, row_major=True):
-        xx = x.repeat(len(y))
-        yy = y.view(-1, 1).repeat(1, len(x)).view(-1)
-        if row_major:
-            return xx, yy
-        else:
-            return yy, xx
-
-    def grid_anchors(self, featmap_size, stride=16, device='cuda'):
-        base_anchors = self.base_anchors.to(device)
-
-        feat_h, feat_w = featmap_size
-        shift_x = torch.arange(0, feat_w, device=device) * stride
-        shift_y = torch.arange(0, feat_h, device=device) * stride
-        shift_xx, shift_yy = self._meshgrid(shift_x, shift_y)
-        shifts = torch.stack([shift_xx, shift_yy, shift_xx, shift_yy], dim=-1)
-        shifts = shifts.type_as(base_anchors)
-        # first feat_w elements correspond to the first row of shifts
-        # add A anchors (1, A, 4) to K shifts (K, 1, 4) to get
-        # shifted anchors (K, A, 4), reshape to (K*A, 4)
-
-        all_anchors = base_anchors[None, :, :] + shifts[:, None, :]
-        all_anchors = all_anchors.view(-1, 4)
-        # first A rows correspond to A anchors of (0, 0) in feature map,
-        # then (0, 1), (0, 2), ...
-        return all_anchors
-
-    def valid_flags(self, featmap_size, valid_size, device='cuda'):
-        feat_h, feat_w = featmap_size
-        valid_h, valid_w = valid_size
-        assert valid_h <= feat_h and valid_w <= feat_w
-        valid_x = torch.zeros(feat_w, dtype=torch.uint8, device=device)
-        valid_y = torch.zeros(feat_h, dtype=torch.uint8, device=device)
-        valid_x[:valid_w] = 1
-        valid_y[:valid_h] = 1
-        valid_xx, valid_yy = self._meshgrid(valid_x, valid_y)
-        valid = valid_xx & valid_yy
-        valid = valid[:, None].expand(
-            valid.size(0), self.num_base_anchors).contiguous().view(-1).bool()
-        return valid
-
-
 class AnchorGeneratorRange(object):
 
     def __init__(self,

mmdet3d/core/bbox/__init__.py

 from . import box_torch_ops
 from .assigners import AssignResult, BaseAssigner, MaxIoUAssigner
-from .coders import ResidualCoder
+from .coders import Residual3DBoxCoder
 # from .bbox_target import bbox_target
 from .iou_calculators import (BboxOverlaps3D, BboxOverlapsNearest3D,
                               bbox_overlaps_3d, bbox_overlaps_nearest_3d)
 from .samplers import (BaseSampler, CombinedSampler,
                        InstanceBalancedPosSampler, IoUBalancedNegSampler,
                        PseudoSampler, RandomSampler, SamplingResult)
-from .transforms import delta2bbox  # bbox2result_kitti,
-from .transforms import (bbox2delta, bbox2result_coco, bbox2roi, bbox_flip,
-                         bbox_mapping, bbox_mapping_back,
-                         boxes3d_to_bev_torch_lidar, distance2bbox, roi2bbox)
+from .transforms import boxes3d_to_bev_torch_lidar
 
 from .assign_sampling import (  # isort:skip, avoid recursive imports
     build_bbox_coder,  # temporally settings
@@ -19,10 +16,9 @@ from .assign_sampling import (  # isort:skip, avoid recursive imports
 __all__ = [
     'BaseAssigner', 'MaxIoUAssigner', 'AssignResult', 'BaseSampler',
     'PseudoSampler', 'RandomSampler', 'InstanceBalancedPosSampler',
-    'IoUBalancedNegSampler', 'CombinedSampler', 'SamplingResult', 'bbox2delta',
-    'delta2bbox', 'bbox_flip', 'bbox_mapping', 'bbox_mapping_back', 'bbox2roi',
-    'roi2bbox', 'bbox2result_coco', 'distance2bbox', 'build_assigner',
-    'build_sampler', 'assign_and_sample', 'box_torch_ops', 'build_bbox_coder',
-    'ResidualCoder', 'boxes3d_to_bev_torch_lidar', 'BboxOverlapsNearest3D',
-    'BboxOverlaps3D', 'bbox_overlaps_nearest_3d', 'bbox_overlaps_3d'
+    'IoUBalancedNegSampler', 'CombinedSampler', 'SamplingResult',
+    'build_assigner', 'build_sampler', 'assign_and_sample', 'box_torch_ops',
+    'build_bbox_coder', 'Residual3DBoxCoder', 'boxes3d_to_bev_torch_lidar',
+    'BboxOverlapsNearest3D', 'BboxOverlaps3D', 'bbox_overlaps_nearest_3d',
+    'bbox_overlaps_3d'
 ]

mmdet3d/core/bbox/assign_sampling.py

@@ -14,7 +14,7 @@ def build_assigner(cfg, **kwargs):
 
 
 def build_bbox_coder(cfg, **kwargs):
-    if isinstance(cfg, coders.ResidualCoder):
+    if isinstance(cfg, coders.Residual3DBoxCoder):
         return cfg
     elif isinstance(cfg, dict):
         return mmcv.runner.obj_from_dict(cfg, coders, default_args=kwargs)

mmdet3d/core/bbox/coders/__init__.py

-from .box_coder import ResidualCoder
+from .box_coder import Residual3DBoxCoder
 
-__all__ = ['ResidualCoder']
+__all__ = ['Residual3DBoxCoder']

mmdet3d/core/bbox/coders/box_coder.py

@@ -2,7 +2,7 @@ import numpy as np
 import torch
 
 
-class ResidualCoder(object):
+class Residual3DBoxCoder(object):
 
     def __init__(self, code_size=7, mean=None, std=None):
         super().__init__()

mmdet3d/core/bbox/samplers/__init__.py

-from .base_sampler import BaseSampler
-from .combined_sampler import CombinedSampler
-from .instance_balanced_pos_sampler import InstanceBalancedPosSampler
-from .iou_balanced_neg_sampler import IoUBalancedNegSampler
-from .ohem_sampler import OHEMSampler
-from .pseudo_sampler import PseudoSampler
-from .random_sampler import RandomSampler
-from .sampling_result import SamplingResult
+from mmdet.core.bbox.samplers import (BaseSampler, CombinedSampler,
+                                      InstanceBalancedPosSampler,
+                                      IoUBalancedNegSampler, OHEMSampler,
+                                      PseudoSampler, RandomSampler,
+                                      SamplingResult)
 
 __all__ = [
     'BaseSampler', 'PseudoSampler', 'RandomSampler',

mmdet3d/core/bbox/samplers/base_sampler.py

-from abc import ABCMeta, abstractmethod
-
-import torch
-
-from .sampling_result import SamplingResult
-
-
-class BaseSampler(metaclass=ABCMeta):
-
-    def __init__(self,
-                 num,
-                 pos_fraction,
-                 neg_pos_ub=-1,
-                 add_gt_as_proposals=True,
-                 **kwargs):
-        self.num = num
-        self.pos_fraction = pos_fraction
-        self.neg_pos_ub = neg_pos_ub
-        self.add_gt_as_proposals = add_gt_as_proposals
-        self.pos_sampler = self
-        self.neg_sampler = self
-
-    @abstractmethod
-    def _sample_pos(self, assign_result, num_expected, **kwargs):
-        pass
-
-    @abstractmethod
-    def _sample_neg(self, assign_result, num_expected, **kwargs):
-        pass
-
-    def sample(self,
-               assign_result,
-               bboxes,
-               gt_bboxes,
-               gt_labels=None,
-               **kwargs):
-        """Sample positive and negative bboxes.
-
-        This is a simple implementation of bbox sampling given candidates,
-        assigning results and ground truth bboxes.
-
-        Args:
-            assign_result (:obj:`AssignResult`): Bbox assigning results.
-            bboxes (Tensor): Boxes to be sampled from.
-            gt_bboxes (Tensor): Ground truth bboxes.
-            gt_labels (Tensor, optional): Class labels of ground truth bboxes.
-
-        Returns:
-            :obj:`SamplingResult`: Sampling result.
-        """
-        bboxes = bboxes[:, :4]
-
-        gt_flags = bboxes.new_zeros((bboxes.shape[0], ), dtype=torch.uint8)
-        if self.add_gt_as_proposals:
-            bboxes = torch.cat([gt_bboxes, bboxes], dim=0)
-            assign_result.add_gt_(gt_labels)
-            gt_ones = bboxes.new_ones(gt_bboxes.shape[0], dtype=torch.uint8)
-            gt_flags = torch.cat([gt_ones, gt_flags])
-
-        num_expected_pos = int(self.num * self.pos_fraction)
-        pos_inds = self.pos_sampler._sample_pos(
-            assign_result, num_expected_pos, bboxes=bboxes, **kwargs)
-        # We found that sampled indices have duplicated items occasionally.
-        # (may be a bug of PyTorch)
-        pos_inds = pos_inds.unique()
-        num_sampled_pos = pos_inds.numel()
-        num_expected_neg = self.num - num_sampled_pos
-        if self.neg_pos_ub >= 0:
-            _pos = max(1, num_sampled_pos)
-            neg_upper_bound = int(self.neg_pos_ub * _pos)
-            if num_expected_neg > neg_upper_bound:
-                num_expected_neg = neg_upper_bound
-        neg_inds = self.neg_sampler._sample_neg(
-            assign_result, num_expected_neg, bboxes=bboxes, **kwargs)
-        neg_inds = neg_inds.unique()
-
-        return SamplingResult(pos_inds, neg_inds, bboxes, gt_bboxes,
-                              assign_result, gt_flags)