search

PINTO0309 / PINTO_model_zoo

A repository for storing models that have been inter-converted between various frameworks. Supported frameworks are TensorFlow, PyTorch, ONNX, OpenVINO, TFJS, TFTRT, TensorFlowLite (Float32/16/INT8), EdgeTPU, CoreML.
https://qiita.com/PINTO
MIT License
3.53k stars 568 forks source link

Explaination on The YoloX Body-Head-Hand output dimension? #381

Closed kho-bluefrogrobotics closed 9 months ago

kho-bluefrogrobotics commented 10 months ago

Issue Type

Documentation Feature Request

OS

Other

OS architecture

armv7

Programming Language

Other

Framework

TensorFlowLite

Model name and Weights/Checkpoints URL

YoloX Body-Head-Hand https://github.com/PINTO0309/PINTO_model_zoo/tree/main/426_YOLOX-Body-Head-Hand

Description

First of all thanks and congratulation for your excellent work, on YOLO.

I have a question regarding the output size, though. @PINTO0309 You say you limit the output to 20 boxes : How come we obtain a float32[60,7] output then? shouldn't we get something like 20x [class, score, x1, y1, x2, y2] = 20 x 5 floats?

Relevant Log Output

No response

URL or source code for simple inference testing code

No response

PINTO0309 commented 9 months ago

As detailed in the README.

max output boxes per class

20 (Body) + 20 (Head) + 20 (Hand) = 60 boxes

[batch_num, classid, score, x1, y1, x2, y2] * 60 = 7 x 60 floats

image

image

kho-bluefrogrobotics commented 9 months ago

OK Thanks!