Welcome to the Data-Centric Track of the Wake Vision Challenge! 🎉
The goal of this track is to push the boundaries of tiny computer vision by enhancing the data quality of the Wake Vision Dataset.
🔗 Learn More: Wake Vision Challenge Details
Participants are invited to:
- Enhance the provided dataset to improve person detection accuracy.
- Train the MCUNet-VWW2 model, a state-of-the-art person detection model, on the enhanced dataset.
- Assess quality improvements on the public test set.
You can modify the dataset however you like, but the model architecture must remain unchanged. 🛠️
First, install Docker on your machine:
Run the following command inside the directory where you cloned this repository:
sudo docker run -it --rm -v $PWD:/tmp -w /tmp andregara/wake_vision_challenge:cpu python data_centric_track.py- This trains the MCUNet-VWW2 model on the original dataset.
- Modify the dataset to improve the model's test accuracy by correcting labels or augmenting data.
💡 Note: The first execution may take several hours as it downloads the full dataset (~365 GB).
- Install the NVIDIA Container Toolkit.
- Verify your GPU drivers.
Run the following command inside the directory where you cloned this repository:
sudo docker run --gpus all -it --rm -v $PWD:/tmp -w /tmp andregara/wake_vision_challenge:gpu python data_centric_track.py- This trains the MCUNet-VWW2 model on the original dataset.
- Modify the dataset to enhance test accuracy while keeping the model architecture unchanged.
💡 Note: The first execution may take several hours as it downloads the full dataset (~365 GB).
- Focus on Data Quality: Explore label correction, data augmentation, and other preprocessing techniques.
- Stay Efficient: The dataset is large, so plan your modifications carefully.
- Collaborate: Join the community discussion on Discord to share ideas and tips!
Have questions or need help? Reach out on Discord.
🌟 Happy Innovating and Good Luck! 🌟