visjs_classificator

Classificator for pictures matching and clustering. Fast and visual.

Uses VISjs and Socket.io to allow image classification/pairing, by hand, with multiuser.

Note : GIF performance is not representative of real time performance.

Launch

node server.js or ./launch_server.sh

If you want to load a previously saved json file :

nodejs server.js .json

Launch (dev)

if you have nodemon or want to develop the tool :

./watch_server.sh

Interaction and usage

CTRL + clic to select more than one node at a time.

After a selection of multiple node, press "M" to create a complete graphe.

Double-clic on a node to make it "fixed" on the screen (your screen only). Perform again to un-fix a node.

Right-clic + drag to draw a selection rectangle

With selection of nodes, press "c" to generate an anchor. If you link a picture to an anchor, you add this node to the cluster.

Interactive thresholding

Performance

Tested up to 5000 pictures at the same time. May still have some tricks to perform to go higher. Depends of your configuration.

Todo list

• Rectangle selection : Clic + CTRL Done, rectangle would still need to be done : Done
• When double clic and moved, the node stays where it is (fixed property isk kept on dragging) : Done
• Simple and double arrows support
• Order on a grid (See Mokkadev repository)
• Selection rectangle : Done
• Increase link size : Done
• Show one or many links between pictures

Good links to get started if you're lost :

https://medium.com/jeremy-keeshin/hello-world-for-javascript-with-npm-modules-in-the-browser-6020f82d1072 https://github.com/jkeesh/hello-browser-npm https://socket.io/docs/emit-cheatsheet/