Computer Vision Internship

Selected intern's day-to-day responsibilities include:

1. Gather data from various sources, including public datasets and real-world captures.
2. Remove corrupt data, handle missing values, and apply augmentation techniques like rotation and scaling.
3. Label data using annotation tools, and normalize and resize images to prepare for model training.
4. Choose appropriate architectures, set up training experiments using PyTorch or TensorFlow, and monitor training metrics.
5. Experiment with hyperparameters and utilize pre-trained models for fine-tuning.
6. Evaluate model performance using validation and test datasets, ensuring robustness through cross-validation.
7. Convert models for deployment on resource-constrained hardware.
8. Apply quantization, pruning, and knowledge distillation to improve model efficiency.
9. Test models on single-board computers (Jetsons, RPIs), addressing latency and throughput issues for real-time applications.
10. Implement techniques like Grad-CAM, LIME, and SHAP to explain model predictions.
11. Create visualizations to explain decisions and document methods and findings.
12. Integrate models with drone and robotic systems, configure cameras, and ensure synchronized data streams.
13. Combine information from multiple sensors for enhanced accuracy.
14. Participate in field tests, capture data and feedback, and refine models and systems based on results.
15. Identify and resolve issues encountered during field tests.
16. Stay updated with the latest research in deep learning, computer vision, and robotics.
17. Conduct experiments, document setups and results, and prepare reports on project progress and findings.

2