search

mystic123 / tensorflow-yolo-v3

Implementation of YOLO v3 object detector in Tensorflow (TF-Slim)
https://medium.com/@pawekapica_31302/implementing-yolo-v3-in-tensorflow-tf-slim-c3c55ff59dbe
Apache License 2.0
890 stars 353 forks source link

Inetegrate letter box image to demo.py #49

Closed i3oi3o closed 5 years ago

i3oi3o commented 5 years ago

I manual test it.

Use the cute doggy image :dog: from here as test image. https://github.com/pjreddie/darknet/blob/master/data/dog.jpg

These are the flag that I use. --data_format=NHWC --size=608 --input_img=dog.jpg --output_img=dog_yolo.jpg

The main change is in utils.convert_to_original_size function .

Let me know if there is anything wrong. :smile:

@sagarkar10 You may want to take a look here. :satisfied:

lucy3589 commented 5 years ago

I manual test it.

Use the cute doggy image 🐶 from here as test image. https://github.com/pjreddie/darknet/blob/master/data/dog.jpg

These are the flag that I use. --data_format=NHWC --size=608 --input_img=dog.jpg --output_img=dog_yolo.jpg

The main change is in utils.convert_to_original_size function .

Let me know if there is anything wrong. 😄

@sagarkar10 You may want to take a look here. 😆

hello, Could you solve it if you modifiy these code(mystic123 approved these changes on 3 Dec 2018)?

i3oi3o commented 5 years ago

@lucy3589 Umm.., Which issue do you want me to resolve?

lucy3589 commented 5 years ago

i use darknet yolo, mAP is 0.9. but i use convert_weight_pb.py to convert weight, and use tf_yolo, but mAP is 0.7.

i3oi3o commented 5 years ago

Unfortunately, I couldn't. This need reverse engineering on both tensorflow and darknet. It's very time consuming. With my current project workload, I don't have time. You can look at #77

lucy3589 commented 5 years ago

Unfortunately, I couldn't. This need reverse engineering on both tensorflow and darknet. It's very time consuming. With my current project workload, I don't have time. You can look at #77

thanks very much. I edit upsample, but it has low mAp.