Home
This graph presents the main characteristics of the dataset.
Yolo v3 have been applied to the ads images (see "Face and object detection" section on the Image Retrieval page. Seven "transports" classes are used: bicycle, car, motorbike, aeroplane, train, truck, boat. Yolo v3 generated 17.5k annotations (1.400 on the means of transport classes).
Yolo may have some serious issues on inferencing objects on heritage newspapers ads. Consequently, a human annotation campaign has been applied to the whole dataset, in order to fix false positives and false negatives. 3,5k annotations have been produced, using the editing features of the GallicaPix web app.