YOLOv5-Pytorch-TensorRT

1 Training in Pytorch

1.1 Download YOLOv5-Pytorch

cd ~/Desktop
mkdir workspace
cd workspace
git clone https://github.com/ultralytics/yolov5

1.2 Prepare dataset

cd ~/Desktop/workspace
mkdir dataset
cd dataset
mkdir annotations
mkdir images
mkdir ImageSets
cd ImageSets
mkdir Main

1.3 Place your images and annotations into images and annotations folder respectively

1.4 Put the split_train_val.py and voc2yolo_label.pyinto dataset folder

1.5 Change voc2yolo_label.py line 8-13 to your config

classes = ["obstacle", "human", "injury"]   # 改成自己的类别
image_dir = "/home/laitathei/Desktop/workspace/dataset/images/"
labels_dir = "/home/laitathei/Desktop/workspace/dataset/labels/"
annotations_dir = "/home/laitathei/Desktop/workspace/dataset/annotations/"
ImageSets_Main_dir = "/home/laitathei/Desktop/workspace/dataset/ImageSets/Main/"
dataset_dir = "/home/laitathei/Desktop/workspace/dataset/"

1.5 Dataset folder structures

├── dataset
   ├── annotations
   │   ├── xxx.xml
   │   
   ├── ImageSets
   │   ├── Main
   │       ├── test.txt
   │       ├── train.txt
   │       ├── trainval.txt
   │       ├── val.txt
   │   
   ├── labels
   │   ├── test
   │       ├── xxx.txt
   |       ├──   ...
   │   ├── val 
   │       ├── xxx.txt
   |       ├──   ...
   │   ├── train
   │       ├── xxx.txt
   |       ├──   ...
   │   
   ├── images
   │   ├── test
   │       ├── xxx.jpg
   |       ├──   ...
   │   ├── val 
   │       ├── xxx.jpg
   |       ├──   ...
   │   ├── train
   │       ├── xxx.jpg
   |       ├──   ...
   ├── train.txt
   ├── train_dummy.txt
   ├── test.txt
   ├── test_dummy.txt
   ├── valid.txt
   ├── valid_dummy.txt
   ├── split_train_val.py
   └── voc2yolo_label.py

1.6 Prepare your dataset yaml file, download the official weight from https://github.com/ultralytics/yolov5/releases/tag/v6.1 and place it into yolov5/weight folder

cd ~/Desktop/workspace/yolov5
mkdir weight
cd weight
wget https://github.com/ultralytics/yolov5/releases/download/v6.1/yolov5n.pt
cd ~/Desktop/workspace/yolov5/data
gedit custom_dataset.yaml

1.7 custom_dataset.yaml content and change your configuration such as number of class and class names

train: /home/laitathei/Desktop/workspace/dataset/train.txt
val: /home/laitathei/Desktop/workspace/dataset/val.txt
test: /home/laitathei/Desktop/workspace/dataset/test.txt
test_xml: /home/laitathei/Desktop/workspace/dataset/annotations

# Classes
nc: 3  # number of classes
names: ['obstacle','human', 'injury']  # class names

1.8 Train your dataset

cd ~/Desktop/workspace/yolov5/
python3 train.py --weights /home/laitathei/Desktop/workspace/yolov5/weight/yolov5n.pt --cfg /home/laitathei/Desktop/workspace/yolov5/models/yolov5n.yaml --data /home/laitathei/Desktop/workspace/yolov5/data/custom_dataset.yaml --epochs 100

1.9 Train result, best.pt and last.pt are FP32 format

100 epochs completed in 0.091 hours.
Optimizer stripped from runs/train/exp/weights/last.pt, 3.9MB
Optimizer stripped from runs/train/exp/weights/best.pt, 3.9MB

Validating runs/train/exp/weights/best.pt...
Fusing layers... 
Model Summary: 213 layers, 1763224 parameters, 0 gradients, 4.2 GFLOPs
               Class     Images     Labels          P          R     mAP@.5 mAP@.5:.95: 100%|██████████| 2/2 [00:00<00:00,  3.06it/s]                                                                      
                 all         55        215      0.986      0.977      0.994      0.916
            obstacle         55        215      0.986      0.977      0.994      0.916
Results saved to runs/train/exp

2 TensorRT conversion

2.1 Convert best.pt->best.wts->best.engine in TensorRTx

2.2 Put the gen_wts.py into yolov5 folder

2.3 Convert best.wts->best.engine by gen_wts.py

cd ~/Desktop/workspace/yolov5
python3 gen_wts.py -w ./runs/train/exp/weights/best.pt -o ./runs/train/exp/weights/best.wts

## Below content will show if program success
YOLOv5 🚀 v6.1-11-g63ddb6f torch 1.10.2+cu113 CPU

2.4 Convert best.wts->best.engine by TensorRTx cmake and best.engine is FP16 format

cd ~/Desktop/workspace/tensorrtx/yolov5
mkdir build
cd build

## Place the best.wts into build folder
## update CLASS_NUM in yololayer.h if your model is trained on custom dataset
## before 
static constexpr int CLASS_NUM = 80; // line 20
static constexpr int INPUT_H = 640;  // line 21  yolov5's input height and width must be divisible by 32.
static constexpr int INPUT_W = 640;  // line 22
## after 
static constexpr int CLASS_NUM = 3; // line 20
static constexpr int INPUT_H = 640; // line 21  yolov5's input height and width must be divisible by 32.
static constexpr int INPUT_W = 640; // line 22

cmake ..
make

## Below content will show if program success
[100%] Linking CXX executable yolov5
[100%] Built target yolov5

sudo ./yolov5 -s best.wts best.engine n
## Below content will show if program success
Loading weights: best.wts
Building engine, please wait for a while...
Build engine successfully!

2.5 Inference the engine with ROS and realsensen D455 camera

2.6 Put inference_ros_trt_v2.py into tensorrtx/yolov5 folder

cd ~/Desktop/workspace/tensorrtx/yolov5
python3 inference_ros_trt_v2.py

## change the realsense camera topic, PLUGIN_LIBRARY, and engine_file_path if required