A multi-step proposal refinement network for improved instance localization and boundary refinement with an application to fish detection

A multi-step proposal refinement network for improved instance localization and boundary refinement with an application to fish detection

We propose a deep learning model that implements a narrow multi-step proposal refining network (MPRN) to increase instance recall and improve horizontal and vertical bound­ary estimation. The MPRN's contribution is a network that learns an automatic correction mechanism from performing step-wis...

Deskribapen osoa

Xehetasun bibliografikoak
Egile nagusia: Labao, Alfonso B. (Egilea)
Beste egile batzuk: Naval, Prospero C. (adviser.)
Formatua: Thesis
Hizkuntza:English
Gaiak:
Computer vision.
Fishes > Detection.
Object localization.