Single-Stage Semantic Segmentation from Image Labels

please download data from origin repo: https://github.com/visinf/1-stage-wseg/tree/master/data

Setup

1. The training requires at least two Titan X GPUs (12Gb memory each).

2. Setup your Python environment. Please, clone the repository and install the dependencies.

   pip install -r requirements.txt
   

3. Download and link to the dataset. We train our model on the original Pascal VOC 2012 augmented with the SBD data (10K images in total). Download the data from:

   • VOC: Training/Validation (2GB .tar file)
   • SBD: Training (1.4GB .tgz file)

   Make sure that the first directory in real_data_path/voc is VOCdevkit; the first directory in real_data_path/sbd is benchmark_RELEASE. in sbd ,you also should change the name cls to cls_png. finally the directory is: ./real_data_path/voc/VOCdevkit/VOC2012/...; ./real_data_path/sbd/benchmark_RELEASE/dataset/...

4. Download pre-trained models. Download the initial weights (pre-trained on ImageNet) for the backbones you are planning to use and place them into <project>/models/weights/.

   Backbone	Initial Weights
   WideResNet38	ilsvrc-cls_rna-a1_cls1000_ep-0001.pth (402M)

Training

To train a model, run train_1P_NPU.py or train_8P_NPU.py with the desired model architecture and the path to the sbd dataset:

# training 1p accuracy
bash ./test/train_full_1P.sh --data_path=real_data_path

# training 1p performance
bash ./test/train_performance_1P.sh --data_path=real_data_path

# training 8p accuracy
bash ./test/train_full_8P.sh --data_path=real_data_path

# training 8p performance 
bash ./test/train_performance_8P.sh --data_path=real_data_path

#test 8p accuracy
bash test/train_eval_8P.sh --data_path=real_data_path --pth_path=real_pre_train_model_path

# finetuning 1p 
bash test/train_finetune_1P.sh --data_path=real_data_path --pth_path=real_pre_train_model_path

Wseg training result

Acc@1	FPS	Npu_nums	Epochs	AMP_Type
-	4.5	1	24	O1
57.0	32	8	24	O1

Statement

For details about the public address of the code in this repository, you can get from the file public_address_statement.md