Improving small objects detection using transformer

General artificial intelligence counteracts the inductive bias of an algorithm and tunes for out-of-distribution generalization. A conspicuous impact is unceasing trend in improving deep learning performance. Although a quintessential attention-based object detection technique, DETR, shows better accuracy than its predecessors, deteriorates detecting small-sized (in-perspective) objects. This study examines DETR proposes normalized using data fusion, SOF-DETR. technique lazy-fusion features introduced SOF-DETR, which sustains contextual information objects present image. The from multiple subsequent layers are fused queries that learn long short-distance spatial association image attention mechanism. Experimental results on MS COCO Udacity Self Driving Car datasets assert effectiveness added feature fusion techniques, showing increased mAP scores