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
3.43k stars 560 forks source link

Difference on model outputs (tflite, openvino IR, and Onnx) in model 227_face-detection-adas-0001 #395

Closed saeidn95 closed 2 months ago

saeidn95 commented 5 months ago

Issue Type

Feature Request

OS

Ubuntu

OS architecture

x86_64

Programming Language

Python

Framework

OpenVINO

Model name and Weights/Checkpoints URL

227_face-detection-adas-0001 https://github.com/PINTO0309/PINTO_model_zoo/tree/main/227_face-detection-adas-0001

Description

I am looking at the outputs of the 3 versions of the model tflite, onnx and IR openvino. The openvino has 3 outputs and ftlite and onnx each have two outputs. How are they identical to each other?

openvino xlm : [<Output: names[mbox_loc] shape[1,40448] type: f32>, <Output: names[mbox_conf_flatten] shape[1,20224] type: f32>, <Output: names[mbox_priorbox] shape[1,2,40448] type: f32>]

tflite: [<Output: names[StatefulPartitionedCall:0] shape[1,40448] type: f32>, <Output: names[StatefulPartitionedCall:1] shape[1,20224] type: f32>]

onnx: [<Output: names[tf.identity] shape[1,40448] type: f32>, <Output: names[tf.identity_1] shape[1,20224] type: f32>]

Relevant Log Output

-

URL or source code for simple inference testing code

-

PINTO0309 commented 2 months ago

I have answered in the past issue. Search for.