search

DevHunter128 / RSSfeed-front

0 stars 0 forks source link

Issue on Image recognition #2

Closed T0pTalent closed 6 months ago

T0pTalent commented 6 months ago

Running into issues with our image recognition AI. It’s misidentifying objects in low-light conditions. Any ideas on improving its accuracy?

BestFriend67 commented 6 months ago

Sounds like it might benefit from a more diverse training set that includes low-light images. Have you tried data augmentation to simulate those conditions?

T0pTalent commented 6 months ago

Not yet, that could work. Do you think enhancing the model with a more robust architecture, like Convolutional Neural Networks (CNN) with attention mechanisms, would help too?

BestFriend67 commented 6 months ago

Absolutely, CNNs excel at image tasks, and adding attention mechanisms can help the model focus on relevant parts of the image, improving performance in challenging conditions. Also, consider fine-tuning with a pre-trained model specifically trained on low-light images.

T0pTalent commented 6 months ago

Great advice! I’ll implement data augmentation for low-light conditions and explore attention-based CNNs. Thanks for the help!

BestFriend67 commented 6 months ago

No problem! Happy to assist. Looking forward to seeing the improvements.