Object detection and counting of tomatoes for further yield estimation by analyzing Unmanned Aerial Vehicles (UAVs) imagery acquired using Low Altitude Remote Sensing (LARS)
- Implemented R-CNN model for estimation of tomatoes in a given tomato farm.
- The model is based on one of the state-of-the-art CNN-based object detection and classification.
- Proposed enhancements over original algorithm :
- Improved accuracy by selecting the hyper parameters from the epoch that has least loss during training
- Refined each bounding box using regression.
Sample frame extracted from the video acquired:
Using labelImg, created the ground truth labels for training(takes a LOT of time):