Gaussian guided IoU: A better metric for balanced learning on object detection

Abstract

Most anchor‐based detectors use intersection over union (IoU) to assign targets to anchors during training. However, IoU did not pay enough attention to the proximity of the anchor's centre to the centre of the truth box, resulting in two issues: (1) the most slender objects were given just one anchor, resulting in insufficient supervision information for slender objects during training; (2) IoU cannot accurately represent the degree of alignment between the feature's receptive field at the anchor's centre and the object. As a result, some features with good alignment degrees are missing, while others with poor alignment degrees are used, reducing the model's localisation accuracy. To address these issues, we first created a Gaussian Guided IoU (GGIoU), which prioritises the proximity of the anchor's centre to the truth box's centre. We then proposed GGIoU‐balanced learning methods, including GGIoU‐guided assignment strategy and GGIoU‐balanced localisation loss. This method can assign multiple anchors to each slender object, favouring features that are well‐aligned with the objects during the training process. A large number of experiments show that GGIoU‐balanced learning can solve the aforementioned problems and significantly improve the detection model's performance.

Multi-person tracking based on spatial location constraint and occlusion awareness

“…We compared the Intersection over Union (IoU), the Distance Intersection over Union (DIoU) [10], and the Generalized Intersection over Union (GIoU) [11]to calculate the occlusion of the target and found that these intersection over union methods are more suitable for target detection, but not very suitable for representing the degree of target occlusion. For the representation of occlusion, unlike the IoU used by algorithms like ImprAsso [12], we use occlusion degree to represent the situation of the target being occluded, that is, the ratio of the occluded area of the target to the total area of the target.…”

Deteksi Denyut Jantung Embrio Ikan Oryzias Celebencis Menggunakan Dekomposisi Kuantil Abu-Abu

“…Sementara itu, data kebenaran dasar region objek bergerak yang harus dideteksi kuantil abu-abu. Selain itu, telah dilakukan pula pengujian pendeteksian objek bergerak memakai pendekatan tumpang tindih area hasil deteksi dan ground truth area yang harus dideteksi per sampel [18]. Berdasarkan pengujian ini, kuantil abu-abu nomor 3 pada sampel ke-1 mendapatkan IoU sebesar 0.85; kuantil abu-abu nomor 4 pada sampel ke-2 mendapatkan IoU sebesar 0.8; kuantil abu-abu nomor 4 pada sampel ke-3 mendapatkan IoU sebesar 0.9; kuantil abu-abu nomor 4 pada sampel ke-4 mendapatkan rata-rata IoU sebesar 0.9; kuantil abu-abu nomor 1 pada sampel ke-5…”