In this work, we have a set of 5 cars for which we want to develop classifiers and detectors.
Collection of cars on the list now:
- Iran Khodro Dena and Dena Plus
- Kia Cerato
- Mazda 3N and Mazda 3
- Peugeot 206
- Saipa Saina
This is done voluntarily. To collect data, car sales websites (e.g. : شیپور) have been used, which have been collected via web crawling.
for classifing bellow models are used :
- my own CNN model
- MobileNetV2 pre-trained model
- YOLO-v4 tiny
for detection bellow model is used:
YOLO-v4 tiny
we have some conditions on our models :
- CNN model can't has more than 2M parameters
- MobileNetV2 model can't has more than 3M parameters
- YOLO model can't has more than 6M parameters