ObjectDetection

This is computer vision project aimed to train models Yolo8 and Detectron2 to detect Fir-tree and Yandex rovers on pictures. In the proccess, after processing all steps and evaluation: Detectron2 model achived more accurate results.

Models used: Yolov8 and Detectron2.

More about models itself you can read on: https://roboflow.com/model/yolov8 and https://roboflow.com/model/detectron2.

Dataset: consists of 104 pictures: Fir-trees and Yandex rovers. Picture size: 6000 x 4000. Resolution: 72 x 72

Dataset itself: https://app.roboflow.com/appliedml/objdetection-uu4zf/1

Step 1: Dataset Collection.

Dataset was colected manually: 100+ photos of fir-tree and yandex rovers was taken.

Step 2: Annotation labels on object using roboflow

Step 3: After data preperation, we push our dataset to detectron2 model: fit and train it.

link to colab: https://colab.research.google.com/drive/1jStZF5V6c29QKw6Bz8EZudsxphBx0VHX?usp=sharing

Step 4: Train Yolov8 in the same manner.

link to colab: https://colab.research.google.com/drive/1BKI7C5Ug7SVN9WJnOpFmfq-TQGeZwyDJ?usp=sharing

Step 5: Evaluation.

Yolov8: detectron2: