<<<<<<< HEAD

We choose Training Data as the sample for training the network, and choose Test B as the validation set to test the ability of our networkk ：Dataset Link

Relevant weights are still being sorted out

After you download the relevant data, put the training set into ./dataset/Train and change the name of the photo folder to JPEGImages

After you download the relevant data, put the Testb set into ./dataset/Test and change the name of the photo folder to JPEGImages

If you want to retrain, execute the following code

export CUDA_VISIBLE_DEVICES=0 
python tools/train.py -c configs/ppyolo/ppyolov2_Bot_50_Xhead_640.yml 

If you have multiple GPUs, execute the following code

export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
python -m paddle.distributed.launch --gpus 0,1,2,3,4,5,6,7 tools/train.py -c configs/ppyolo/ppyolov2_Bot_50_Xhead_640.yml

If you want to perform an evaluation, you can execute

export CUDA_VISIBLE_DEVICES=0 
python tools/eval.py -c configs/ppyolo/ppyolov2_Bot_50_Xhead_640.yml -o weights={path of Weights}

We use PaddleDetection as the basic framework to build PaddleDetection