Because your footage is near perfect, you won't be disappointed, but try from a different angle, or from a bit greater distance and CodeProject fails, while Plate Recognizer gets results even from an angle where you can hardly see it yourself anymore: Screenshot 2024-12-26 171412.jpg Analyzed with ...

Remove the check from this

Try these settings

pack2010 wrote: ↑Sun Dec 22, 2024 4:31 pm 82 I think, the metrics had started to level off at that point.

I find you need at least 200 epochs, Is there any way you can share the dataset and I will try training it. I have RTX 4090 and it shouldn't take that long to train the model.

I've gotten all the images (at least 500 for each of the 36 species I'm working with), is there a good auto labeler that's not exorbitantly expensive? Roboflow limits the amount of images I can upload before I have to start spending quite a bit of money. You can use Make Sense to label your images,...

I've gotten all the images (at least 500 for each of the 36 species I'm working with), is there a good auto labeler that's not exorbitantly expensive? Roboflow limits the amount of images I can upload before I have to start spending quite a bit of money. You can use Make Sense to label your images,...

Below is the list of supported GPU's by generation for the ALPR module

OK, looking good. For me it went like this: 14 Dec 2024 Installing CPAI 2.9.5 over 2.6.5 1. Download CPai 2.9.5 from https://codeproject.github.io/ 2. Download latest NVIDIA driver for GTX1650. 3. Download latest CUDA driver. 4. Download latest cuDNN script. 5. Disable BI5 service. Uninstall CPAI (...

I retrained the model using the Roboflow image. The model does work but not well because of the image count. On average there is only 40 images for each type of bird. To have a better model you need at a minimum 500 images of each type of bird. Below are links to the model and training results. http...

Can you post the link to the model on Roboflow and I will test it on my system. If the model was trained using YOLOv5 it should work with YOLOv6.2