Environments

1. Python 3.6
2. Pytorch 0.4.0
3. $HOME=TrajectoryDrawing

Preparation

1. install python3.6
2. install pytorch 0.4.0
   • 1. download whl file "cu90/torch-0.4.0-cp36-cp36m-linux_x86_64.whl" from https://pytorch.org/get-started/previous-versions/
   • 2. pip3 install torch_XXX.whl
3. cd ${HOME}/faster_rcnn.pytorch
4. pip3 install requirements.txt

Directory hierarchy

TrajectoryDrawing/
+ faster_rcnn.pytorch/
+ demo_video/
+ Moon_morning_2_1.mp4
+ Moon_morning_2_3.mp4
+ data/
+ pretrained_model/
+ vgg16_caffe.pth (pretrained vgg16 for training and testing faster-RCNN backend)
+ models/
+ vgg16/
+ car_type_0/
+ faster_rcnn_1_50_150.pth (trained faster-RCNN model, used for predicting)
+ prepare_demo_frames.sh (將影片分割成frame，然後放進適當的資料夾)
+ test_net.py (predict每個frame，預測bbox)
+ output_video.sh (產生有bbox的影片到./demo_video中)
+ visualize_movement.py (產生trajectory的影片到./demo_video中)
+ dataset/
+ car_type_0/
+ results/
+ demo_type_0/
+ data/
+ Images/
+ (prepare_demo_frames.sh output的frame們都會放在這)
+ ImageSets/
+ (prepare_demo_frames.sh 產生的demo_type_0.txt會放在這，文件包含所有要被test的frame)

Dataset type

1. type 0: 俯視
2. type 1: 測俯視

Execution

1. type 0

   model name: models/vgg16/car_type_0/faster_rcnn_1_50_150.pth

   ./prepare_demo_frames.sh
   CUDA_VISIBLE_DEVICES=2 python3 test_net.py --dataset car_type_0 --net vgg16 --checksession 1 --checkepoch 50 --checkpoint 150 --cuda
   ./output_video.sh
   python3 visualize_movement.py --video 1 --start 600 --end 1050 --object pedestrian --object_idx 26 --output_dir ./demo_video
   python3 visualize_movement.py --video 2 --start 960 --end 1080 --object bicycle --object_idx 0 --output_dir ./demo_video

2. type 1

   model name: models/vgg16/car_type_1/faster_rcnn_1_50_128.pth

   CUDA_VISIBLE_DEVICES=2 python3 test_net.py --dataset car_type_1 --net vgg16 --checksession 1 --checkepoch 50 --checkpoint 128 --cuda