helps to analyse integrated circuit die images (for example from siliconpr0n.org) with the help of ai.
Project description
Installation
Install from source:
pip install .
Install from pip:
pip install silicon-analyser
Information
Code will use your graphic card for acceleration. (but only if correct pytorch is installed, see "Additional info" below)
Frameworks/Libraries used:
Small example
- start
- select image
- add grid
- press mouse down on image and drag your rectangle for your grid
- adjust x,y,cols,rows,width,height manualy to fit
- add label (while grid is selected)
- give it a random name
- with that label selected, select cells for that label (for example cells that mark a "1")
- select grid (for example "grid_0" again)
- add another label (while grid is selected)
- give it a random name
- with that label selected, select cells for that label (for example cells that mark a "0")
- with enough "1" and "0" labels drawn, click the "Compute" button
- ai will find images in the grid that have the same properties
- click "stop" once the results are satisfied
- maximum for "acc" and "val_acc" is 1.00, the closer you are to those values, the better are the results
- results depend on many factors:
- the amount of cells you selected
- how good your grid matches the current image
- the quality of your image
- ...
- "acc" stands for "accuracy", "val" for "validation"
- found ai-cells will be drawn green
Additional info
- you might need to install cuda-specific PyTorch for accelerated computing
- check your graphic driver version for compatible cuda version!
Keys
- Use up/down/left/right to navigate
- Hold shift to move faster
- Scroll-wheel to zoom out
- Click on minimap to get directly to a position
- Right click on tree-items (left navigation menu) for additional options
TODO
- show loading screen on start (pytorch with cuda support takes a bit to load)
- auto-compute to calculate in background while you are selecting new cells for your labels
- ai-model configuration
- possibility to rotate grid
- maybe store your model on a public place? (for others to use)
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
silicon_analyser-1.0.6.tar.gz
(25.1 kB
view hashes)
Built Distribution
Close
Hashes for silicon_analyser-1.0.6-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 6dee1b05b0cfd298ddb493ceff28653070aa712fbac7485ebd54bf423709bfdf |
|
MD5 | 000223b32c1f0f950d9169552bfded16 |
|
BLAKE2b-256 | 14a32dc52f9d6bcd65b54ecc1cc49f297866b20626539428c24037e8521ad960 |