Egg line anomaly detection, now running locally 🥚 with no cloud API calls:
@erik_kokalj1.4K followers
14,032 views
14 Comments
Comments from the original post on X (Twitter)

@TrevorHLynn
@roboflow how does the model handle glare?
3 likes

@erik_kokalj
@roboflow hm good question - if anomaly det model was trained on only eggs with no glare, it'd likely flag glare as an anomaly. So one would need to add egg images with glare to training set, that way, model would learn that glare is not an anomaly
2 likes

@markfenner
@roboflow This is pretty cool
1 like

@erik_kokalj
@roboflow thanks Mark 🫡
1 like

@markfenner
@roboflow Any time Erik
1 like

@chainlockw
@roboflow 用的什么芯片

@erik_kokalj
@roboflow nvidia jetson, runs at about 110ms per anomaly det inference, can be optimized with batching tho!

@chainlockw
@roboflow 这个看起来很酷,我想学习
1 like

@aklen
@roboflow @ClarenceCourt 👀

@erik_kokalj
@roboflow @ClarenceCourt 👀

@moueza
@roboflow Eggs intelligence

@Spring_Hermit
@roboflow 所以这个鸡蛋背面也能拍到吗?反转的时候不会碎吗?

@erik_kokalj
@roboflow it's a rotating conveyor belt, so the whole shell of the egg is inspected
1 like

@moueza
@roboflow Dynamic Segmentation + XR