Muhammad Shoaib Manzoor

Academic

Career

Verbatim text

VOLA 

Visual Object Labelling Assistant, is a web bases dataset labelling tool where multiple labelers and 
reviewers can collaborate on dataset labelling and creation process. Key stockholders are labelers, 
reviewers, and clients. This semi-automated collocative process begins when client uploads dataset 
onto our server, selects labelers and reviewers, and approve the end product of VOLVA, the 
labelled dataset. Initiation: In this stage, client uploads archive file of unlabeled dataset onto our 
server where it’s uncompressed, divided into segments when required which usually the case with 
satellite images because of its huge dimensions, and labelers and reviewers are assigned to the 
dataset labelling process task. Labeling : Soon as labelers get allocated portion of the dataset 
assigned by reviewer, they can begin annotation process. Each assigned segment appears in the 
workspace to labeler one by one, showing progress bar above. Labeler can move back and forth 
through segments using navigation buttons. Once segment appears in workspace having objects of 
interests, the labeler starts new instances, selects bounding boxes and respective labels for those 
instances, and move onto next image segment. This goes on until whole assigned portion gets 
labelled. During this process, the labeler can also take suggestions from assistive model by clicking 
‘show suggestions’ button which is greyed out initially because of insufficient labelled images. After 
enough labelled instances, the ‘show suggestions’ button appears in workspaces indicating that 
labeler can begin speeding up task assigned with AI assistance.Reviewing: Once annotated by 
labelers, these images are passed on to reviewers who monitor labelers work. If satisfied they can 
approve annotation and finish this process, and if not, they can either fix annotations themselves or 
reassigned those incorrect annotations to labelers who are responsible for that task.  




























Technology Used: 
Django, MongoDb, React 
Tensorflow 
Supervisor Name: 
Mr. Hassan Mustafa 
Group Members:   
Rehman Gul (i16 - 0306) 
Muhammad Shoaib Manzoor (i16 - 0109) 
Zubair Shahid (i16 - 0081)