DOIONLINE 
DOIONLINE Publish In |
International Journal of Advances in Electronics and Computer Science-IJAECS |
 Journal Home Volume Issue |
||||||||
Issue |
Volume-9,Issue-12 ( Dec, 2022 ) | |||||||||
Paper Title |
Recursive Label Refinement for Weakly Semi Supervised Animal Parsing using CNN | |||||||||
Author Name |
Karthik S, Adithya Babu, Venu Madhav Nookala, Abhilash Sk | |||||||||
Affilition |
1,2,3Software Engineer, KPIT Technologies, Bangalore, India 4Senior Tech Lead, KPIT Technologies, Bangalore, India | |||||||||
Pages |
21-29 | |||||||||
Abstract |
Abstract - While segmentation of objects is an ongoing evolving technique used for different class objects, the pixel level part based segmentation of objects remains an exigent task in this domain. Labeling pixel level masks for all object classes is cumbersome and error prone. Hence in this paper a Recursive Label Refinement Animal Parsing (RLRAP) technique is implemented using a self correcting parsing network. Applying this model promotes the authenticity of the learned models as well as the supervised labels .The RLRAP technique uses the iterative cycles for segmenting parts of unseen object classes given that they have similar structures as seen objects. In a nutshell, an Iterative Learning Scheduler (ILS) is leveraged to recursively aggregate the current learned model with the previous optimum one in order to infer more credible pseudo- masks. In this approach, during the cycles of ILS, both the models and the labels will progressively become more robust and accurate. RLRAP has been benchmarked on the standard animal parsing datasets Pascal Part Segmentation to infer that this model thrives to outperform the existing state of the art architectures. Keywords - Parts Segmentation, Animal Parsing, Recursive Label Refinement, Iterative Learning Scheduler (ILS) | |||||||||
| View Paper | ||||||||||
