Description
Weapon detector prototype for Engineering Company. The main solution was to classify weapons like short guns and knives more accurately. Challenge was provide professional AI Computer Vision solution for weapon detection accuracy greater than 70%.
For this project we optimized deep learning trainings by adding synthetic data (dataset) and distributed training with TensorFlow and 8 Nvidia Tesla V100 GPUs on S3 instances in Amazon Web Service (AWS).
We achieved training of a dataset with 300.000 models in just 40 minutes for this mobile prototype.
Object Tracking and Clasiffication
Technologies Used
- TensorFlow
- Phyton
- Blender
- Amazon Web Services (AWS)
- C++
- Android Studio
