vision_cuda.h

#pragma once
#if defined(WITH_CUDA)
#include <c10/cuda/CUDAGuard.h>
#elif defined(WITH_HIP)
#include <c10/hip/HIPGuard.h>
#endif
#include <torch/extension.h>

at::Tensor ROIAlign_forward_cuda(
    const at::Tensor& input,
    const at::Tensor& rois,
    const float spatial_scale,
    const int pooled_height,
    const int pooled_width,
    const int sampling_ratio,
    const bool aligned);

at::Tensor ROIAlign_backward_cuda(
    const at::Tensor& grad,
    const at::Tensor& rois,
    const float spatial_scale,
    const int pooled_height,
    const int pooled_width,
    const int batch_size,
    const int channels,
    const int height,
    const int width,
    const int sampling_ratio,
    const bool aligned);

std::tuple<at::Tensor, at::Tensor> ROIPool_forward_cuda(
    const at::Tensor& input,
    const at::Tensor& rois,
    const float spatial_scale,
    const int pooled_height,
    const int pooled_width);

at::Tensor ROIPool_backward_cuda(
    const at::Tensor& grad,
    const at::Tensor& rois,
    const at::Tensor& argmax,
    const float spatial_scale,
    const int pooled_height,
    const int pooled_width,
    const int batch_size,
    const int channels,
    const int height,
    const int width);

std::tuple<at::Tensor, at::Tensor> PSROIPool_forward_cuda(
    const at::Tensor& input,
    const at::Tensor& rois,
    const float spatial_scale,
    const int pooled_height,
    const int pooled_width);

at::Tensor PSROIPool_backward_cuda(
    const at::Tensor& grad,
    const at::Tensor& rois,
    const at::Tensor& mapping_channel,
    const float spatial_scale,
    const int pooled_height,
    const int pooled_width,
    const int batch_size,
    const int channels,
    const int height,
    const int width);

std::tuple<at::Tensor, at::Tensor> PSROIAlign_forward_cuda(
    const at::Tensor& input,
    const at::Tensor& rois,
    const float spatial_scale,
    const int pooled_height,
    const int pooled_width,
    const int sampling_ratio);

at::Tensor PSROIAlign_backward_cuda(
    const at::Tensor& grad,
    const at::Tensor& rois,
    const at::Tensor& mapping_channel,
    const float spatial_scale,
    const int pooled_height,
    const int pooled_width,
    const int sampling_ratio,
    const int batch_size,
    const int channels,
    const int height,
    const int width);

at::Tensor nms_cuda(
    const at::Tensor& dets,
    const at::Tensor& scores,
    const float iou_threshold);

at::Tensor DeformConv2d_forward_cuda(
    const at::Tensor& input,
    const at::Tensor& weight,
    const at::Tensor& offset,
    const at::Tensor& bias,
    std::pair<int, int> stride,
    std::pair<int, int> pad,
    std::pair<int, int> dilation,
    int groups,
    int deformable_groups);

std::tuple<at::Tensor, at::Tensor, at::Tensor, at::Tensor>
DeformConv2d_backward_cuda(
    const at::Tensor& grad_out,
    const at::Tensor& input,
    const at::Tensor& weight,
    const at::Tensor& offset,
    const at::Tensor& bias,
    std::pair<int, int> stride,
    std::pair<int, int> pad,
    std::pair<int, int> dilation,
    int groups,
    int deformable_groups);