[cherry-pick]fix problem due to empty bbox_pred and modify ppyolo docs, test=dygraph (#2180)

* fix problem due to empty bbox_pred and modify ppyolo docs, test=dygraph

* modify post process to avoid dy2st problem

dygraph/configs/ppyolo/README.md

@@ -46,7 +46,7 @@ PP-YOLO improved performance and speed of YOLOv3 with following methods:
 | PP-YOLO_2x               |     8      |     24     | ResNet50vd |     512     |         44.4         |         45.0          |      89.9      |          188.4          | [model](https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ppyolo_r50vd_dcn_2x_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/dygraph/configs/ppyolo/ppyolo_r50vd_dcn_2x_coco.yml)                   |
 | PP-YOLO_2x               |     8      |     24     | ResNet50vd |     416     |         42.7         |         43.2          |      109.1      |          215.4          | [model](https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ppyolo_r50vd_dcn_2x_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/dygraph/configs/ppyolo/ppyolo_r50vd_dcn_2x_coco.yml)                   |
 | PP-YOLO_2x               |     8      |     24     | ResNet50vd |     320     |         39.5         |         40.1          |      132.2      |          242.2          | [model](https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ppyolo_r50vd_dcn_2x_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/dygraph/configs/ppyolo/ppyolo_r50vd_dcn_2x_coco.yml)                   |
-| PP-YOLO_ResNet18vd               |     4      |     32     | ResNet18vd |     512     |         29.3         |         29.5          |      357.1      |          657.9          | [model](https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ppyolo_r18vd_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/dygraph/configs/ppyolo/ppyolo_r18vd_coco.yml)                   |
+| PP-YOLO_ResNet18vd               |     4      |     32     | ResNet18vd |     512     |         29.2         |         29.5          |      357.1      |          657.9          | [model](https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ppyolo_r18vd_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/dygraph/configs/ppyolo/ppyolo_r18vd_coco.yml)                   |
 | PP-YOLO_ResNet18vd               |     4      |     32     | ResNet18vd |     416     |         28.6         |         28.9          |      409.8      |          719.4          | [model](https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ppyolo_r18vd_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/dygraph/configs/ppyolo/ppyolo_r18vd_coco.yml)                   |
 | PP-YOLO_ResNet18vd               |     4      |     32     | ResNet18vd |     320     |         26.2         |         26.4          |      480.7      |          763.4          | [model](https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ppyolo_r18vd_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/dygraph/configs/ppyolo/ppyolo_r18vd_coco.yml)                   |
 
@@ -129,7 +129,7 @@ CUDA_VISIBLE_DEVICES=0 python tools/infer.py configs/ppyolo/ppyolo_r50vd_dcn_1x_
 CUDA_VISIBLE_DEVICES=0 python tools/infer.py configs/ppyolo/ppyolo_r50vd_dcn_1x_coco.yml -o weights=https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ppyolo_r50vd_dcn_1x_coco.pdparams --infer_dir=../demo
 ```
 
-### 4. Inferece deployment and benchmark
+### 4. Inferece deployment
 
 For inference deployment or benchmard, model exported with `tools/export_model.py` should be used and perform inference with Paddle inference library with following commands:
 
@@ -141,18 +141,6 @@ python tools/export_model.py -c configs/ppyolo/ppyolo_r50vd_dcn_1x_coco.yml -o w
 CUDA_VISIBLE_DEVICES=0 python deploy/python/infer.py --model_dir=output_inference/ppyolo_r50vd_dcn_1x_coco --image_file=../demo/000000014439_640x640.jpg --use_gpu=True
 ```
 
-Benchmark testing for PP-YOLO uses model without data reading and post-processing(NMS), export model with `--exclude_nms` to prunce NMS for benchmark testing from mode with following commands:
-
-```bash
-# export model, --exclude_nms to prune NMS part, model will be save in output/ppyolo as default
-python tools/export_model.py -c configs/ppyolo/ppyolo_r50vd_dcn_1x_coco.yml -o weights=https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ppyolo_r50vd_dcn_1x_coco.pdparams --exclude_nms
-
-# FP32 benchmark
-CUDA_VISIBLE_DEVICES=0 python deploy/python/infer.py --model_dir=output_inference/ppyolo_r50vd_dcn_1x_coco --image_file=../demo/000000014439_640x640.jpg --use_gpu=True --run_benchmark=True
-
-# TensorRT FP16 benchmark
-CUDA_VISIBLE_DEVICES=0 python deploy/python/infer.py --model_dir=output_inference/ppyolo_r50vd_dcn_1x_coco --image_file=../demo/000000014439_640x640.jpg --use_gpu=True --run_benchmark=True --run_mode=trt_fp16
-```
 
 ## Future work
 

dygraph/configs/ppyolo/README_cn.md

@@ -46,7 +46,7 @@ PP-YOLO从如下方面优化和提升YOLOv3模型的精度和速度：
 | PP-YOLO_2x               |     8      |     24     | ResNet50vd |     512     |         44.4         |         45.0          |      89.9      |          188.4          | [model](https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ppyolo_r50vd_dcn_2x_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/dygraph/configs/ppyolo/ppyolo_r50vd_dcn_2x_coco.yml)                   |
 | PP-YOLO_2x               |     8      |     24     | ResNet50vd |     416     |         42.7         |         43.2          |      109.1      |          215.4          | [model](https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ppyolo_r50vd_dcn_2x_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/dygraph/configs/ppyolo/ppyolo_r50vd_dcn_2x_coco.yml)                   |
 | PP-YOLO_2x               |     8      |     24     | ResNet50vd |     320     |         39.5         |         40.1          |      132.2      |          242.2          | [model](https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ppyolo_r50vd_dcn_2x_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/dygraph/configs/ppyolo/ppyolo_r50vd_dcn_2x_coco.yml)                   |
-| PP-YOLO_ResNet18vd               |     4      |     32     | ResNet18vd |     512     |         29.3         |         29.5          |      357.1      |          657.9          | [model](https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ppyolo_r18vd_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/dygraph/configs/ppyolo/ppyolo_r18vd_coco.yml)                   |
+| PP-YOLO_ResNet18vd               |     4      |     32     | ResNet18vd |     512     |         29.2         |         29.5          |      357.1      |          657.9          | [model](https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ppyolo_r18vd_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/dygraph/configs/ppyolo/ppyolo_r18vd_coco.yml)                   |
 | PP-YOLO_ResNet18vd               |     4      |     32     | ResNet18vd |     416     |         28.6         |         28.9          |      409.8      |          719.4          | [model](https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ppyolo_r18vd_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/dygraph/configs/ppyolo/ppyolo_r18vd_coco.yml)                   |
 | PP-YOLO_ResNet18vd               |     4      |     32     | ResNet18vd |     320     |         26.2         |         26.4          |      480.7      |          763.4          | [model](https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ppyolo_r18vd_coco.pdparams) | [config](https://github.com/PaddlePaddle/PaddleDetection/tree/master/dygraph/configs/ppyolo/ppyolo_r18vd_coco.yml)                   |
 
@@ -128,7 +128,7 @@ CUDA_VISIBLE_DEVICES=0 python tools/infer.py configs/ppyolo/ppyolo_r50vd_dcn_1x_
 CUDA_VISIBLE_DEVICES=0 python tools/infer.py configs/ppyolo/ppyolo_r50vd_dcn_1x_coco.yml -o weights=https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ppyolo_r50vd_dcn_1x_coco.pdparams --infer_dir=../demo
 ```
 
-### 4. 推理部署与benchmark
+### 4. 推理部署
 
 PP-YOLO模型部署及推理benchmark需要通过`tools/export_model.py`导出模型后使用Paddle预测库进行部署和推理，可通过如下命令一键式启动。
 
@@ -140,18 +140,6 @@ python tools/export_model.py -c configs/ppyolo/ppyolo_r50vd_dcn_1x_coco.yml -o w
 CUDA_VISIBLE_DEVICES=0 python deploy/python/infer.py --model_dir=output_inference/ppyolo_r50vd_dcn_1x_coco --image_file=../demo/000000014439_640x640.jpg --use_gpu=True
 ```
 
-PP-YOLO模型benchmark测试为不包含数据预处理和网络输出后处理(NMS)的网络结构部分数据，导出模型时须指定`--exlcude_nms`来裁剪掉模型中后处理的NMS部分，通过如下命令进行模型导出和benchmark测试。
-
-```bash
-# 导出模型，通过--exclude_nms参数裁剪掉模型中的NMS部分，默认存储于output/ppyolo目录
-python tools/export_model.py -c configs/ppyolo/ppyolo_r50vd_dcn_1x_coco.yml -o weights=https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ppyolo_r50vd_dcn_1x_coco.pdparams --exclude_nms
-
-# FP32 benchmark测试
-CUDA_VISIBLE_DEVICES=0 python deploy/python/infer.py --model_dir=output_inference/ppyolo_r50vd_dcn_1x_coco --image_file=../demo/000000014439_640x640.jpg --use_gpu=True --run_benchmark=True
-
-# TensorRT FP16 benchmark测试
-CUDA_VISIBLE_DEVICES=0 python deploy/python/infer.py --model_dir=output_inference/ppyolo_r50vd_dcn_1x_coco --image_file=../demo/000000014439_640x640.jpg --use_gpu=True --run_benchmark=True --run_mode=trt_fp16
-```
 
 ## 未来工作
 

dygraph/ppdet/modeling/post_process.py

@@ -36,6 +36,10 @@ class BBoxPostProcess(object):
         else:
             bbox_pred, bbox_num = self.decode(head_out, rois, im_shape,
                                               scale_factor)
+        if bbox_pred.shape[0] == 0:
+            bbox_pred = paddle.to_tensor(
+                np.array(
+                    [[-1, 0.0, 0.0, 0.0, 0.0, 0.0]], dtype='float32'))
         return bbox_pred, bbox_num
 
     def get_pred(self, bboxes, bbox_num, im_shape, scale_factor):
@@ -51,7 +55,7 @@ class BBoxPostProcess(object):
                                bboxes are corresponding to the original image.
         """
         if bboxes.shape[0] == 0:
-            return paddle.to_tensor([[0, 0.0, 0.0, 0.0, 0.0, 0.0]])
+            return paddle.zeros(shape=[1, 6])
 
         origin_shape = paddle.floor(im_shape / scale_factor + 0.5)