I am continuing to improve data collection with my web application myML.
The reason for this feature is to group all the images to the model being built. Even when the focus on a campaign or project. If you are building a model on a subject or object, it would be true anyone else doing the same.
The gallery that I am using is an open-source PHP gallery call folioGallery. Click on the link, I may make some adaptions to ensure its error-free for my setup. Currently using the old format of folioGallery that I found, but will make an adjustment to the current version since it seems to support video.
All images being stored here have a bounding box xy minimum and xy maximum points. Image resolutions are approximately 400×400 pixels to make easier to upload on a mobile network.
The figure above, show how the bounding boxes compare to each other. The one person doing the captures can best select the bounding box that would match the subject. The dashed line represents the bounding box.
This enable to concentrate on a single model and classification labeling campaign.
Since it uses the browser that supports HTML 5, the tech stack also includes webRTC and makes it compatible with any device.
My thoughts in using this application are to make it available as a service from my website. I think for it to be successful, it needs to be in a controlled environment. Otherwise, you have a lot of folks uploading images and video that has nothing to do with the target subject and model.