fatih akyon

anladigim kadariyla yolox kutuphanesi icin ekstra resize islemi yapmaya gerek yok, resizei kendi icinde hallediyor: https://github.com/Megvii-BaseDetection/YOLOX/blob/74b637b494ad6a968c8bc8afec5ccdd7ca6b544f/tools/demo.py#L140-%20L146 image_size degerini test_size olarak set etmemiz yeterli olur gibi anliyorum: https://github.com/Megvii-BaseDetection/YOLOX/blob/74b637b494ad6a968c8bc8afec5ccdd7ca6b544f/tools/demo.py#L118

@kadirnar source baktim, yolox projesindeki hubconfta size hardcoded set edilmis. bence ugrasmaya gerek yok, kullanici yolox icin 640 disinda bir size set etmeye calisirsa yoloxdetectionmodel image_size set etmeyi desteklemiyor gibi...

@kadirnar firsat buldugunda main branchteki yenilikleri rebase ya da merge ile alip devam edebilirsin calismalarina

> @fcakyon YOLOX ve YOLOV7 modellerini ayrı ayrı oluşturmak yerine https://github.com/open-mmlab/mmyolo direk bu kütüphaneyi ekleyebiliriz. evet mantikli, yolov7 destegi eklemeyi dusunuyolar mi diye issue acilabilir o repoya 👍

@Rushikesh-pawar yes you can definitely use. You have to add the support yourself. You can follow the steps in main readme under contributing section 👍

@ppmzhang2 there is no need for this update. pycocotools package has windows support.

@SkalskiP In this fork, I have added roboflow support for ultralytics==8.0.35. https://github.com/fcakyon/ultralytics/tree/roboflow Also added continuous integration for roboflow support: https://github.com/fcakyon/ultralytics/actions/runs/4156300655

> @fcakyon you plan to maintain that fork? ;) I have been maintaining it for three weeks. Not guaranteeing constant support, but I will update in my free time :)...

@Robotatron dataset scanning seems to be faster with `v5loader=True`.

Actually, I have only tested with `v5loader=True`. Let me try with `v5loader=False`.