kneron / model_zoo

Kneron AI Training Platform Model_Zoo

Introduction

We provide a collection of classification and detection models pre-trained on the ImageNet dataset and the COCO dataset. In the table below, we summarized each such pre-trained model including:

• a model name.
• model input size.
• model speed: we report frame per second (fps) evaluated on our 520 and 720 hardwares.
• model size.
• model performance on the ImageNet validation set and COCO validation set.

Pre-trained Classification Models

Model	Input Size	FPS on 520	FPS on 720	Model Size	Rank 1 Accuracy	Rank 5 Accuracy
mobilenetv2	224x224	58.9418	620.677	14M	69.82%	89.29%
resnet18	224x224	20.4376	141.371	46.9M	66.46%	87.09%
resnet50	224x224	6.32576	49.0828	102.9M	72.80%	90.91%
FP_classifier	56x32	323.471	3370.47	5.1M	94.13%	-

mobilenetv2, resnet18 and resnet50 are models pre-trained on ImageNet classification dataset. FP_classifier is a model pre-trained on our own dataset for classifying person and background images.

Resnet50 is currently under training for Kneron preprocessing.

Pre-trained Detection Models

Backbone	Input Size	FPS on 520	FPS on 720	Model Size	mAP
YOLOv5s (no upsample)	640x640	4.91429	-	13.1M	40.4%
YOLOv5s (with upsample)	640x640	-	24.4114	14.6M	50.9%
FCOS (darknet53s backbone)	416x416	7.27369	48.8437	33.9M	44.8%

model_zoo's Issues

a few precisions on the FP_classifier model?

Hello,

I tried to use your provided human clasifier model (greatly apreciated, btw, Thanks! )

However, I obtain weird and unstable results.

Could you give some insights on how to use this model?

I guessed that the first output is for the background recognition, and the second for the humans, right?
What are the expected output values?
I guess we need to normalise the input image, with what mean and std_var values?
Does it work with humans that are seated, cropped, horizontal?

Thanks for your answer!

PS: I'm using the OpenCV dnn engine
which looks like smthing like that (Java):

             Net humanClassifier;
             humanClassifier = Dnn.readNet("./FP_classifier.onnx");
             Mat blob = Dnn.blobFromImage(inputImage, 1/255,
                    new org.opencv.core.Size(32, 56),
                    new Scalar(new double[]{0.0, 0.0, 0.0}), /*swapRB*/false, /*crop*/false, CV_32F);
            humanClassifier.setInput(blob);

            List<Mat> outputs = new ArrayList<Mat>();
            humanClassifier.forward(outputs);