0

I have downloaded this source.

https://github.com/Austin-Ellsworth/Vehicle_Recognition_System/find/main

For Vehicle detection + Make/Model Recognition.

The Make/Model Recognition part is done using TensorFlow/MobileNetsV3. The Weights.pb,Labels, Config file are available.

But no Checkpoint files. And Also not aware of the dataset they used to train the model.

So is it possible to evaluate this classifier model with these files only? (To aquire maP for example?...)

If not, will evaluating it using a different dataset give me an accurate maP?

And what is the most efficient way of doing it?

Thanks

Mandroid
  • 1

1 Answers1

1

*.pb files stands for protobuf. Protobuf is used by tensorflow to save model graph and weights.

Model is actualy a YOLOv3 ported to tensorflow running on original YOLOv3 weights. It was trained at COCO dataset so it will give you similar accuracy on this dataset.

kacpo1
  • 525
  • 5
  • 17
  • There are 2 weights in the folder the .pb is actually for "MobileNetsV3" The YOLO Model Weights are Located in the folder YOLO-COCO. This project Consists of both detection using YOLO, and Vehicle Model Classification with Mobilenets/TensorFlow – Mandroid Jun 28 '21 at 11:21