As we have experimented Detectron 2, there is one more leap of faith which remaining

I might have mentioned about YOLOv5 in the YOLO part, If you watch marvel movies, you become a fan of the MARVEL universe. But you can’t deny the fact that you’re already a fan of DC universe as like if you use Detectron 2 with the R-CNN family, You could try as a fan of Yolov5.

So, the flow is same as the previous part

1. Upload the images to the Roboflow

2. Instead of generating the dataset with COCO format, We can use YOLOv5 PyTorch to generate. In our case, the dataset remains the same, to incorporate the comparison between our models

3. The Training Code is attached blow – Run the model

Please wait a minute. Where is the accuracy?

The reason is the model was trained for 100epochs and the training time is 6mins -> 😲That’s incredibly fast, but then I have altered our model with 5000epochs and guess what!

This image has an empty alt attribute; its file name is yolo_inference.jpg
The training time was double that of Detectron 2

4. Like we have seen in YOLOv3, testing the images with the last weights and best weights -> I have used the same strategy here

Best model weight results
Last model weight results

Not that much change, both were good, but the Best one is slightly low if you consider accuracy.

That’s all about YOLOv5 and implementation

In the next part, let’s put an end card to Object Detection😂We’ll conclude by comparing both the models which we have seen in the multi-class detection

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s