tensorflow 1.x not working anymore

[mehrajagdish]

please update the notebook as colab removed the support for tensorflow 1.x
this doesn't work anymore%tensorflow_version 1.x

[Erol444]

@mehrajagdish which notebook?

[mehrajagdish]

@Erol444 https://github.com/luxonis/depthai-ml-training/blob/master/colab-notebooks/Easy_Object_Detection_Demo_Training.ipynb https://github.com/luxonis/depthai-ml-training/blob/master/colab-notebooks/Easy_Object_Detection_With_Custom_Data_Demo_Training.ipynb

[Erol444]

@tersekmatija Could someone from ML team look into updating these mobilenet-ssd training notebooks?

[HonzaCuhel]

Hello, I apologize for the slight delay in my answer. We have looked at the notebooks and ultimately decided to deprecate them, since the notebooks are rather old, the Yolo models (e.g., yolov6n) offer better and more accurate performance, and we can easily export the Yolo model weights to a .blob format using our tools. We would like to invite you to check out our Colab notebooks on training YoloV5, YoloV6, YoloV7. Thank you for understanding, Jan

[mehrajagdish]

@HonzaCuhel but yolo models are heavy compared to mobilenet and hence while running yolo models i am getting low fps.

[HonzaCuhel]

Hi, Not necessarily. Nano versions of YoloV5 and YoloV6 models are small and run fast on OAK cameras (for example, yolov6n runs around 50 FPS on OAK-1). I would recommend using the yolov6n model, for which we have prepared a Colab notebook link to see how to train such a model. Best, Jan