Hi,
We are a thrift store and receive a lot of clothes.
We want to use Vision to recognize the brand of our clothes.
80% of the time, it works without any problem, but the rest of the time, it makes error. Sometimes it mispells a brand, sometime it create new ones.
We would like to find the best way to make it work at 100%.
What would be the best way to do it ?
Thanks,
you can always train your object detection object as a model trained to recognize clothes is most definitely going to be more accurate, you can also mess with the top p and temperature settings to try to get closer and run it through 3 times and take the most common result.
Do you have enough ground truth examples of the brands + images? If so you could just finetune a small classifier on top of pretrained image embeddings.