sarthak268 / Animal-Detection

Detection of Animals in camera trapped images using RetinaNet in pytorch.

Animal-Detection

Link for saved weights

• Commands for evaluating the model on heldout test set.

mkdir Test_Images

All the images will be placed in this folder

mkdir Test_Images_Final 

This folder will contain all the images after data cleaning and converting the annotations.json to annotations_test.csv.

Keep the annotations.json file in the same directory as the clean_data.py.

• For testing, first clean the test data:

python clean_data.py

• Command for testing Baseline Model:

python test_baseline.py --test_anno_file annotations_test.csv --type baseline

• Command for testing the Baseline Model with Ensemble Approach:

python test_baseline.py --test_anno_file annotations_test.csv --type baseline_ensemble

• Command for testing the Baseline Model with Weighted Random Sampling:

python test_improved.py --test_anno_file annotations_test.csv --type improved

• Command for testing the Baseline Model with Weighted Random Sampling and Ensemble Approach:

python test_improved.py --test_anno_file annotations_test.csv --type improved_ensemble

• Command for visualising the Baseline Model:

python visualize.py --dataset csv --csv_classes classname2id.csv --csv_val annotations_test.csv --model <saved_model_pth>

• Command for visualising results with Weighted Random Sampling:

python visualize.py --dataset csv --csv_classes classname2id.csv --csv_val annotations_test.csv --model <saved_model_pth>

• Command for visualising results with Ensemble Approach:

python visualize.py --dataset csv --csv_classes classname2id.csv --csv_val annotatoins_test.csv --model <saved_model_pth>

Team

• Anish Madan
• Sarthak Bhagat
• Shagun Uppal