Downsize first FITS file more on EXOTIC Colab in Step 3

[rzellem]

Some large files (see attached) do not load in Step 3 (where you ID your target and comp stars) - so we need to downsample the first image more to ensure that it loads and does not time out the Colab
Uploading 2024-01-29_20-02-35_UVIR_10.00_Gain100_Focus-24136_60.00s_1x1_0012_c_20240324212540.fits.zip…

[ivenzor]

@tamimfatahi @rzellem, I’ve worked a bit on this issue. Before creating a PR, could you take a look and let me know if this aligns with what you have in mind? I’m attaching an image (I tried to attach the HTML file, but it’s not supported).

New Features:

1. Downsampling the image to 1MB to balance speed and resolution if the input image exceeds 1MB.
2. When hovering, the original coordinates are shown instead of the downsampled ones.
3. Added controls to flip the image horizontally and vertically while retaining the original coordinates.
4. Added controls to adjust how the image is displayed (low percentile and high percentile settings).

I could also add RA and DEC coordinates when hovering, provided the image is plate solved.

[rzellem]

@ivenzor - these changes look awesome!

Can you talk more about the downsampling? How does it do that? Does it bin the pixel size? Or something else?

If it bins the pixel size, I'd be a little worried that it could "bin away" some of the smaller/dimmer stars.

[ivenzor]

@rzellem currently I'm using the skimage.transform.resize function. It's a spline interpolation, Bi-linear by default:

0: Nearest-neighbor
1: Bi-linear (default)
2: Bi-quadratic
3: Bi-cubic
4: Bi-quartic
5: Bi-quintic
Default is 0 if image.dtype is bool and 1 otherwise.

[rzellem]

Very cool! I didn't know about this feature- this is super helpful! I wonder if we should also build this into the Image Inspector as well?

[ivenzor]

I’ll work on a PR for the 'locating target star' step, and if everything goes well, I’ll also take a look at the notebook code for inspecting the images.