VehicleAttributeRecognitionBarrier : A Machine Learning Model for Detecting Car Attributes

This is an introduction to「VehicleAttributeRecognitionBarrier」, a machine learning model that can be used with ailia SDK. You can easily use this model to create AI applications using ailia SDK as well as many other ready-to-use ailia MODELS.

Overview

VehicleAttributeRecognitionBarrier is a machine learning model developed by Intel to identify the type and the color of a car.

open_model_zoo/README.md

Source: https://pixabay.com/ja/videos/%E8%AD%A6%E5%AF%9F%E3%81%AE%E8%BB%8A-%E5%B8%82-%E3%83%88%E3%83%A9%E3%83%95%E3%82%A3%E3%83%83%E3%82%AF-6095/

Architecture

The model takes a frontal image of a car and outputs attributes.

Source: https://github.com/openvinotoolkit/open_model_zoo/blob/master/models/intel/vehicle-attributes-recognition-barrier-0042/README.md

There are 7 categories for colors and 4 categories for vehicle types, with accuracies of respectively 82.71% and 87.34%

The model architecture has a ResNet-like structure. The input is a (1,72,72,3) image and the output is a (1,7) vector of color probabilities and a (1,4) vector of car types.

Source: https://netron.app/?url=https://storage.googleapis.com/ailia-models/vehicle-attributes-recognition-barrier/vehicle-attributes-recognition-barrier-0042.onnx.prototxt

As a constraint, you need to give the image of the front of the car with less than 50% occlusion.

Usage

You can use VehicleAttributeRecognitionBarrier with ailia SDK using the following command. After detecting the car with YOLOv3 for any video, the attributes will be inferred.

$ python3 vehicle-attributes-recognition-barrier.py -v input.mp4

Here is an example output.

ailia-models/vehicle_recognition/vehicle-attributes-recognition-barrier

ax Inc. has developed ailia SDK, which enables cross-platform, GPU-based rapid inference.

ax Inc. provides a wide range of services from consulting and model creation, to the development of AI-based applications and SDKs. Feel free to contact us for any inquiry.