Two pre-built models are provided, one to identify the make (based on resnet34) and the other to identify the model (based on resnet50) of the vehicle in a supplied image. Both models are residual convolution networks. The training dataset contains over 136,000 tagged images, and 163 car makes with 1,716 car models, not all possible car makes and models are included. Thus, random images of cars may or may not be accurately identified. With a test set of some 2,000 cars the accuracy was found to be in the high 90%.
A GPU with more than 8GB of video memory is recommended for training the model on your own collections of labelled images. A guide to building a deep learning workstation is available to build your own machine, using an Asus NVidia GForce RTX 2080 Ti GPU with 64GB RAM.
Training a model is an iterative process and depends on experience. Each iteration (experiment) might take a few hours of compute time on a GPU, training with 10 epochs. Tuning will occur so as to find the optimal model.
The car makes represetned in the training dataset include: ABT, BAC, Conquest, DS, Dacia, Fisker, GMC, Gumpert, Hennessey, Icona, Jeep, KTM, MELKUS, MG, MINI, Mazzanti, Noble, PGO, SPIRRA, SSC, Scion, TESLA, TVR, Tramontana, Zenvo, smart, Yiqi, Mitsubishi, Shangqidatong, Spyker N.V., Dongnan, Dongfeng, Dongfengxiaokang, Dongfengfengdu, Dongfengfengshen, Dongfengfengxing, Zxauto, Zhonghua, Toyota, Zinoro, Jiulong, Isuzu, Wuling, AC Chnitzer, Zoyte, Iveco, Bufori, Porsche, Mitsuoka, Chrysler, Lamorghini , Kombat, Cadillac, Buck, Lifan, Lorinser, Rolls-Royce, BAW, Baihc, Beiqiweiwang, Beiqihuansu, Beiqi New Energy, Huapu, Huatai, Huaqi, Carlsson, Shuanghuan, Shuanglong, Geely, Venucia, Haval, Hafei, Volkswagen, Daihatsu, Chrey, Besturn, Benz, Audi, Wisemann, Wealeak, BWM, Baojun, Bentley, Brabus, Bugatti, Pagani, Guangqichuanqi, GAC, Karry, Ciimo, CHTC, Jaguar, Morgan, Subaru, Skoda, Xinkai, Nissan, Changhe, RANZ, Honda, Lincoln, Peugeot, Opel, Oley, BYD, Jonway, Huizhong, Jianghuai, Jiangling, Vauxhall, Volvo, Ferrari, Haige, Haima, Haima(Zhengzhou), Cheetah, Maserati, Hyundai , Everus, Ruiqi, Fuqiqiteng, Ford, Futian, Fudi, Koenigsegg, HongQi, Luxgen, SAAB, Denza, Yingzhi, Infiniti, Roewe, Lotus, FIAT, Saab, Lancia, Seat, Qoros, Acura, KIA, Lotus, LAND-ROVER, McLaren, Maybach, Dodge, Mustang, Jinlv, Jinbei, Suzuki, GreatWall, Changan Business, Changan, Alfa Romeo, Aston Martin, Lufeng, Shanqitongjia, Chevy, Citroen, Lexus, Renault, Shouwang, MAZDA, and Huanghai.
Further resources include:
Your donation will support ongoing availability and give you access to the PDF version of this book. Desktop Survival Guides include Data Science, GNU/Linux, and MLHub. Books available on Amazon include Data Mining with Rattle and Essentials of Data Science. Popular open source software includes rattle, wajig, and mlhub. Hosted by Togaware, a pioneer of free and open source software since 1984. Copyright © 1995-2022 Graham.Williams@togaware.com Creative Commons Attribution-ShareAlike 4.0