Based on a fast neural network architecture, our car make and model recognition module can be easily integrated into applications that require accurate tagging of car images. It is robust under different lighting conditions and different angles. It can operate on embedded hardware, on-premise servers or can be deployed as cloud API.
There are lots of industries that could benefit from our car recognition module, including security, marketing and law enforcement. Some sample use cases are:
Intelligent Video Surveillance
Tagging of Video and Images
License Plate Verification
The car make and model classifier that we offer is just a binary neural network model in TensorFlow format. There is no object detector included, and the developers have to use their own vehicle detector to find the cars in each frame. The detected cars must be cropped, padded to square images, and resized to 224x224 pixels, which is the input image size of the classifier. The car classifier is based on MobileNet neural network architecture. It is very fast and runs in real time on CPU of a regular PC. One car image classification takes 35 milliseconds on Intel Core i5-7600 CPU. For faster inference a NVIDIA GPU is recommended to be used. We have also a light version of the classifier with slightly lower accuracy but 4 times faster. There are many ways to integrate the car classifier into your software. For example, the model can be opened in OpenCV by DNN module. It is also possible to use TensorFlow library and to run the classifier using C++ or Python. Another option is to use TensorFlow Serving, which is a high-performance serving system for machine learning models, designed for production environments. It exposes RESTful API (in port 8501) and gRPC interface (in port 8500). The model server can be packaged in Docker container and to be hosted on the cloud or On-Premises servers.
The car make and model classifier has several versions with different input sizes: 224px; 128px; 96px; 64px. The price of the different versions is available here. The 64x64 version is free and can be used without limitations.
The web demo of the 224x224 versions can be tried here.
Number of supported car brands: 216
Number of supported car models: 2902
Minimum vehicle size: 30x30 pixels
Accuracy: 75% - 95%, depending on the dataset
Speed: 35 milliseconds on Intel Core i5-7600 CPU
View angles: front, rear, side view
Recommended open-source object detectors with state-of-art accuracy:
SSD detector with MobileNet V2 feature extractor
Single Shot Multibox Detector (SSD) with Inception V2 feature extractor
Faster R-CNN with Inception Resnet v2 feature extractor