zznidar/colony-counter

Automatically count colonies from a photo of your Petri dish

colony-counter's Issues

Filter matches by height/width ratio

Colonies are usually not linear, but rather circular. By getting the "colony" width/height ratio, we can discard lines, letters and some other noise.

Make this an offline PWA

But include a sharing link somewhere – because from a standalone app the browser chrome is not visible.

Alternatively, use a different webapp type (but make sure it still works offline).

Uncaught TypeError: a[(sredina + i)] is undefined

Should a colony centre (I guess) be found near the image edges, this line gets index out of range.

colony-counter/index.js

Line 507 in 246c6cb

a[sredina + i][y + j] = -1; // already checked

example image: 271035948_679357069896602_9078540942867362065_n.jpg
sredina = 766, i = 2, sredina+i = 768, but resized image size is only 768x1024, thus indices 0:767 and 0:1023

Instead of hard-coded -3:5, we should do something that takes image size into account.

Analysis of large images takes tremendously long time

A picture taken directly with phone camera is usually quite big. There are mindestens three possible speed-up options:

• Use webworkers and split the canvas among multiple threads for analysis
• Resize the image to make it smaller (bacteria collonies are supposedly larger than camera pixels)
• Skip loops outside of circle-bounding square altogether for(let x = centre - r; x < centre + r; x++)

Red colour sometimes spills over the dish, counting "colonies" outside the dish

The green circle shows the area which should be analysed.

Actually, there's a fix in the comment already 😅

colony-counter/index.js

Line 456 in 2bbec04

// TODO: while y < y_end to prevent spilling over the dish. Check it doesn't produce false-positives then.

Avoid counting Petri dish edges

Instead of setting neighbouring pixels to 0 when reaching the centre of a colony, set them to 2 (or them+1) if they > 0.
When reaching the next centre, only count it if the value == 1. Should the value be > 1, do not count the centre, but still set the neighbours to 2.

This should avoid counting continuous small crossings that appear at circle edges.

Black square is always black on iOS

Apparently iOS does not support changing the colour of the medium black square (◼). Thus, the detected/selected background colour always appears black.

Gradient over agar

Depending on the lightning when capturing a photo of your petri dish, a (usually linear) gradient of brightness may appear over the agar. Bacteria on one side might be darker than agar on the other side (or vice-versa).

In such cases, it would be useful to be able to define a gradient of the background, against which the colour should be checked.

Maybe this will be more user-friendly once auto-dish-detection and auto-agar-colour detection are implemented.