Training Sample Solution for LPCVC 2023

The sample solution is based on FANet (Hu, et al. "Real-time semantic segmentation with fast attention", IEEE RA-L, 2021).

0. Installation:

Environment:

1. Linux
2. Python 3.7
3. Pytorch 1.8
4. NVIDIA GPU + CUDA 10.2

Build

pip install -r requirements.txt

1. Prepare Data

Download and save the training/validation data G-Drive (Please send an access request with your team's registriation information.)

2. Modify Codes

Modify *.yml files in ./config

• data:path: path to dataset
• training:batch_size: batch_size
• training:train_augmentations:rcrop: input size for training

3. Train

Run

python train.py --config configs/*.yml

Taining log for the sample solution is provided in sample_solution/runs/FA_Res18/86059/run_2023_01_18_17_48_22.log

4. Validation

Modify model path validating:resume in ./config/*.yml

Run

python val.py --config configs/*.yml