Add YOLOv5 support for RV1126 device.

configs/_base_/backends/rknn.py

@@ -1,8 +1,10 @@
 backend_config = dict(
     type='rknn',
     common_config=dict(
-        mean_values=None,  # [[103.53, 116.28, 123.675]],
-        std_values=None,  # [[57.375, 57.12, 58.395]],
         target_platform='rv1126',  # 'rk3588'
         optimization_level=1),
-    quantization_config=dict(do_quantization=False, dataset=None))
+    quantization_config=dict(
+        do_quantization=True,
+        dataset=None,
+        pre_compile=False,
+        rknn_batch_size=-1))

configs/mmcls/classification_rknn-fp16_static-224x224.py

@@ -0,0 +1,7 @@
+_base_ = ['./classification_static.py', '../_base_/backends/rknn.py']
+
+onnx_config = dict(input_shape=[224, 224])
+codebase_config = dict(model_type='end2end')
+backend_config = dict(
+    input_size_list=[[3, 224, 224]],
+    quantization_config=dict(do_quantization=False))

configs/mmcls/classification_rknn_static-224x224.py → configs/mmcls/classification_rknn-int8_static-224x224.py

@@ -1,5 +1,5 @@
 _base_ = ['./classification_static.py', '../_base_/backends/rknn.py']
 
 onnx_config = dict(input_shape=[224, 224])
-codebase_config = dict(model_type='rknn')
+codebase_config = dict(model_type='end2end')
 backend_config = dict(input_size_list=[[3, 224, 224]])

configs/mmdet/detection/detection_rknn-fp16_static-320x320.py

@@ -0,0 +1,34 @@
+_base_ = ['../_base_/base_static.py', '../../_base_/backends/rknn.py']
+
+onnx_config = dict(input_shape=[320, 320])
+
+codebase_config = dict(model_type='rknn')
+
+backend_config = dict(
+    input_size_list=[[3, 320, 320]],
+    quantization_config=dict(do_quantization=False))
+
+# # yolov3, yolox for rknn-toolkit and rknn-toolkit2
+# partition_config = dict(
+#     type='rknn',  # the partition policy name
+#     apply_marks=True,  # should always be set to True
+#     partition_cfg=[
+#         dict(
+#             save_file='model.onnx',  # name to save the partitioned onnx
+#             start=['detector_forward:input'],  # [mark_name:input, ...]
+#             end=['yolo_head:input'],  # [mark_name:output, ...]
+#             output_names=[f'pred_maps.{i}' for i in range(3)])  # out names
+#     ])
+
+# # retinanet, ssd, fsaf for rknn-toolkit2
+# partition_config = dict(
+#     type='rknn',  # the partition policy name
+#     apply_marks=True,
+#     partition_cfg=[
+#         dict(
+#             save_file='model.onnx',
+#             start='detector_forward:input',
+#             end=['BaseDenseHead:output'],
+#             output_names=[f'BaseDenseHead.cls.{i}' for i in range(5)] +
+#             [f'BaseDenseHead.loc.{i}' for i in range(5)])
+#     ])

configs/mmdet/detection/detection_rknn_static-320x320.py → configs/mmdet/detection/detection_rknn-int8_static-320x320.py

@@ -6,24 +6,27 @@
 
 backend_config = dict(input_size_list=[[3, 320, 320]])
 
-# yolov3, yolox
+# # yolov3, yolox for rknn-toolkit and rknn-toolkit2
 # partition_config = dict(
 #     type='rknn',  # the partition policy name
 #     apply_marks=True,  # should always be set to True
 #     partition_cfg=[
 #         dict(
 #             save_file='model.onnx',  # name to save the partitioned onnx
 #             start=['detector_forward:input'],  # [mark_name:input, ...]
-#             end=['yolo_head:input'])  # [mark_name:output, ...]
+#             end=['yolo_head:input'],  # [mark_name:output, ...]
+#             output_names=[f'pred_maps.{i}' for i in range(3)])  # out names
 #     ])
 
-# # retinanet, ssd, fsaf
+# # retinanet, ssd, fsaf for rknn-toolkit2
 # partition_config = dict(
 #     type='rknn',  # the partition policy name
 #     apply_marks=True,
 #     partition_cfg=[
 #         dict(
 #             save_file='model.onnx',
 #             start='detector_forward:input',
-#             end=['BaseDenseHead:output'])
+#             end=['BaseDenseHead:output'],
+#             output_names=[f'BaseDenseHead.cls.{i}' for i in range(5)] +
+#             [f'BaseDenseHead.loc.{i}' for i in range(5)])
 #     ])

configs/mmdet/detection/yolov3_partition_onnxruntime_static.py

@@ -8,5 +8,6 @@
         dict(
             save_file='yolov3.onnx',
             start=['detector_forward:input'],
-            end=['yolo_head:input'])
+            end=['yolo_head:input'],
+            output_names=[f'pred_maps.{i}' for i in range(3)])
     ])

configs/mmpose/pose-detection_rknn-fp16_static-256x192.py

@@ -0,0 +1,10 @@
+_base_ = ['./pose-detection_static.py', '../_base_/backends/rknn.py']
+
+onnx_config = dict(input_shape=[192, 256])
+
+codebase_config = dict(model_type='end2end')
+
+backend_config = dict(
+    input_size_list=[[3, 256, 192]],
+    quantization_config=dict(do_quantization=False),
+    common_config=dict(target_platform='rv1126'))

configs/mmpose/pose-detection_rknn-fp16_static-256x256.py

@@ -0,0 +1,10 @@
+_base_ = ['./pose-detection_static.py', '../_base_/backends/rknn.py']
+
+onnx_config = dict(input_shape=[256, 256])
+
+codebase_config = dict(model_type='end2end')
+
+backend_config = dict(
+    input_size_list=[[3, 256, 256]],
+    quantization_config=dict(do_quantization=False),
+    common_config=dict(target_platform='rv1126'))

configs/mmpose/pose-detection_rknn-int8_static-256x192.py

@@ -0,0 +1,9 @@
+_base_ = ['./pose-detection_static.py', '../_base_/backends/rknn.py']
+
+onnx_config = dict(input_shape=[192, 256])
+
+codebase_config = dict(model_type='end2end')
+
+backend_config = dict(
+    input_size_list=[[3, 256, 192]],
+    common_config=dict(target_platform='rv1126'))

configs/mmpose/pose-detection_rknn-int8_static-256x256.py

@@ -0,0 +1,9 @@
+_base_ = ['./pose-detection_static.py', '../_base_/backends/rknn.py']
+
+onnx_config = dict(input_shape=[256, 256])
+
+codebase_config = dict(model_type='end2end')
+
+backend_config = dict(
+    input_size_list=[[3, 256, 256]],
+    common_config=dict(target_platform='rv1126'))

configs/mmpose/pose-detection_simcc_rknn-fp16_static-256x192.py

@@ -0,0 +1,7 @@
+_base_ = ['./pose-detection_static.py', '../_base_/backends/rknn.py']
+
+onnx_config = dict(input_shape=[192, 256], output_names=['simcc_x', 'simcc_y'])
+
+backend_config = dict(
+    input_size_list=[[3, 256, 192]],
+    quantization_config=dict(do_quantization=False))

configs/mmpose/pose-detection_simcc_rknn-int8_static-256x192.py

@@ -0,0 +1,5 @@
+_base_ = ['./pose-detection_static.py', '../_base_/backends/rknn.py']
+
+onnx_config = dict(input_shape=[192, 256], output_names=['simcc_x', 'simcc_y'])
+
+backend_config = dict(input_size_list=[[3, 256, 192]])

configs/mmseg/segmentation_rknn-fp16_static-320x320.py

@@ -0,0 +1,9 @@
+_base_ = ['./segmentation_static.py', '../_base_/backends/rknn.py']
+
+onnx_config = dict(input_shape=[320, 320])
+
+codebase_config = dict(model_type='rknn')
+
+backend_config = dict(
+    input_size_list=[[3, 320, 320]],
+    quantization_config=dict(do_quantization=False))

csrc/mmdeploy/codebase/mmdet/base_dense_head.cpp

@@ -81,7 +81,8 @@ Result<Detections> BaseDenseHead::GetBBoxes(const Value& prep_res, const Tensor&
 
     MMDEPLOY_DEBUG("{}-th box: ({}, {}, {}, {}), {}, {}", i, x1, y1, x2, y2, label_id, score);
 
-    auto rect = MapToOriginImage(x1, y1, x2, y2, scale_factor.data(), 0, 0, ori_width, ori_height);
+    auto rect =
+        MapToOriginImage(x1, y1, x2, y2, scale_factor.data(), 0, 0, ori_width, ori_height, 0, 0);
     if (rect[2] - rect[0] < min_bbox_size_ || rect[3] - rect[1] < min_bbox_size_) {
       MMDEPLOY_DEBUG("ignore small bbox with width '{}' and height '{}", rect[2] - rect[0],
                      rect[3] - rect[1]);

csrc/mmdeploy/codebase/mmdet/object_detection.cpp

@@ -6,6 +6,7 @@
 #include "mmdeploy/core/utils/device_utils.h"
 #include "mmdeploy/core/utils/formatter.h"
 #include "mmdeploy/experimental/module_adapter.h"
+#include "utils.h"
 
 using namespace std;
 
@@ -127,6 +128,13 @@ Result<Detections> ResizeBBox::GetBBoxes(const Value& prep_res, const Tensor& de
     scale_factor = {1.f, 1.f, 1.f, 1.f};
   }
 
+  int top_padding = 0;
+  int left_padding = 0;
+  if (prep_res.contains("pad_param")) {
+    top_padding = prep_res["pad_param"][0].get<int>();
+    left_padding = prep_res["pad_param"][1].get<int>();
+  }
+
   float w_offset = 0.f;
   float h_offset = 0.f;
   if (prep_res.contains("border")) {
@@ -153,7 +161,7 @@ Result<Detections> ResizeBBox::GetBBoxes(const Value& prep_res, const Tensor& de
     MMDEPLOY_DEBUG("ori left {}, top {}, right {}, bottom {}, label {}", left, top, right, bottom,
                    *labels_ptr);
     auto rect = MapToOriginImage(left, top, right, bottom, scale_factor.data(), w_offset, h_offset,
-                                 ori_width, ori_height);
+                                 ori_width, ori_height, top_padding, left_padding);
     if (rect[2] - rect[0] < min_bbox_size_ || rect[3] - rect[1] < min_bbox_size_) {
       MMDEPLOY_DEBUG("ignore small bbox with width '{}' and height '{}", rect[2] - rect[0],
                      rect[3] - rect[1]);
@@ -170,15 +178,6 @@ Result<Detections> ResizeBBox::GetBBoxes(const Value& prep_res, const Tensor& de
   }
   return objs;
 }
-std::array<float, 4> ResizeBBox::MapToOriginImage(float left, float top, float right, float bottom,
-                                                  const float* scale_factor, float x_offset,
-                                                  float y_offset, int ori_width, int ori_height) {
-  left = std::max(left / scale_factor[0] + x_offset, 0.f);
-  top = std::max(top / scale_factor[1] + y_offset, 0.f);
-  right = std::min(right / scale_factor[2] + x_offset, (float)ori_width - 1.f);
-  bottom = std::min(bottom / scale_factor[3] + y_offset, (float)ori_height - 1.f);
-  return {left, top, right, bottom};
-}
 
 MMDEPLOY_REGISTER_CODEBASE_COMPONENT(MMDetection, ResizeBBox);
 

csrc/mmdeploy/codebase/mmdet/object_detection.h

@@ -22,10 +22,6 @@ class ResizeBBox : public MMDetection {
   template <typename T>
   Result<Detections> GetBBoxes(const Value& prep_res, const Tensor& dets, const Tensor& labels);
 
-  std::array<float, 4> MapToOriginImage(float left, float top, float right, float bottom,
-                                        const float* scale_factor, float x_offset, float y_offset,
-                                        int ori_width, int ori_height);
-
   std::vector<Tensor> GetDetsLabels(const Value& prep_res, const Value& infer_res);
 
  protected:

csrc/mmdeploy/codebase/mmdet/utils.cpp

@@ -10,11 +10,12 @@ namespace mmdeploy::mmdet {
 
 std::array<float, 4> MapToOriginImage(float left, float top, float right, float bottom,
                                       const float* scale_factor, float x_offset, float y_offset,
-                                      int ori_width, int ori_height) {
-  left = std::max(left / scale_factor[0] + x_offset, 0.f);
-  top = std::max(top / scale_factor[1] + y_offset, 0.f);
-  right = std::min(right / scale_factor[2] + x_offset, (float)ori_width - 1.f);
-  bottom = std::min(bottom / scale_factor[3] + y_offset, (float)ori_height - 1.f);
+                                      int ori_width, int ori_height, int top_padding,
+                                      int left_padding) {
+  left = std::max((left - left_padding) / scale_factor[0] + x_offset, 0.f);
+  top = std::max((top - top_padding) / scale_factor[1] + y_offset, 0.f);
+  right = std::min((right - left_padding) / scale_factor[2] + x_offset, (float)ori_width - 1.f);
+  bottom = std::min((bottom - top_padding) / scale_factor[3] + y_offset, (float)ori_height - 1.f);
   return {left, top, right, bottom};
 }
 

csrc/mmdeploy/codebase/mmdet/utils.h

@@ -11,7 +11,8 @@
 namespace mmdeploy::mmdet {
 std::array<float, 4> MapToOriginImage(float left, float top, float right, float bottom,
                                       const float* scale_factor, float x_offset, float y_offset,
-                                      int ori_width, int ori_height);
+                                      int ori_width, int ori_height, int top_padding,
+                                      int left_padding);
 // @brief Filter results using score threshold and topk candidates.
 // scores (Tensor): The scores, shape (num_bboxes, K).
 // probs: The scores after being filtered