Assignment 1: Image Classifier Tutorial#143
Conversation
1. Worked in the rst file to put a site with tutorial 2. Added New Species such as flamingos and eagles to tutorial 3. Added some comments for full visualization of table to tutorial
| @@ -0,0 +1,337 @@ | |||
| Image Classifier to PyReason Tutorial | |||
There was a problem hiding this comment.
Add the tutorial to index.rst so it shows up in the docs.
There was a problem hiding this comment.
I added the following line at the index.rst
./image_classifier_reasoning.rst
I must admit, for some reason it is not showing up when going to the other tutorial pages, but it does appear in the general one.
| This produces explanations for which rule fired, on which node/edge, and why. | ||
|
|
||
|
|
||
| Expected Output (Full Run + Analysis) |
There was a problem hiding this comment.
Don't show the full classifier output here - highlight the important steps for each image.
There was a problem hiding this comment.
I decided to divide the traces per animals and make a general explanation that when an animal is identified it is set to [1,1]
| Fixed Point iterations: 3 | ||
| NODE RULE TRACE: | ||
|
|
||
| Time Fixed-Point-Operation Node Label Old Bound New Bound Occurred Due To Consistent Triggered By Inconsistency Message Clause-1 |
There was a problem hiding this comment.
Instead of just pasting the full trace, highlight key groundings from each timestep and explain why they were grounded.
There was a problem hiding this comment.
I decided to put what happens on each pass trough and a summarized run of the trace by covering the tables with ...
| NODE RULE TRACE: | ||
|
|
||
| Time Fixed-Point-Operation Node Label Old Bound New Bound Occurred Due To Consistent Triggered By Inconsistency Message Clause-1 | ||
| 0 0 0 fish_1 goldfish [0.0,1.0] [1.0,1.0] fish_1-goldfish-fact True Fact None |
There was a problem hiding this comment.
Add the full trace in examples/csv_outputs
There was a problem hiding this comment.
Csv files have been created based on a run.
1. Improved reading documentation 2. Divided the output more effectively 3. Created Csv Files with output 4. Added the tutorial to the index
1. Removed additional traces 2. Added images of expected output 3. Added source of LLM into tutorial
2 participants
Changes
examples/image_classifier_ex.py— standalone example scriptdocs/source/tutorials/image_classifier_reasoning.rst— tutorial documentation pageexamples/images/— sample images for flamingos and eaglesSummary
Builds an image-classification reasoning example with fish, sharks, flamingos, and eagles.
Uses a pretrained ViT model to classify local images, converts predictions into PyReason facts, and runs rule-based inference to derive:
is_fish,is_shark,is_flamingo,is_eagle,is_birdcan_flyfromis_birdlikes_to_eatedges (shark/flamingo -> fish)Also includes a full tutorial page explaining the pipeline (image -> classifier -> facts -> reasoning -> output), expected terminal output, and warnings about using only labels that exist in the model.
Terminal Output