Update to work with python 3.7+ requirements.txt

INSTRUCTIONS.md

@@ -0,0 +1,59 @@
+<h1>Setup</h1>
+To begin setup.<br><br>
+
+1. I forked the original Caxton repo:<br>
+<a href="https://github.com/cam-cambridge/caxton">https://github.com/cam-cambridge/caxton</a><br><br>
+
+2. I cloned my "Caxton_Fork" to `gebru` using:<br>
+<code>git clone git@github.com:philip-hub/caxton_fork.git</code><br><br>
+
+3. I created a virtual environment:<br>
+<code>python3 -m venv env</code><br>
+I then activated my environment:<br>
+<code>source env/bin/activate</code><br><br>
+
+4. I ran the command to install requirements:<br>
+<code>pip install requirements.txt</code><br>
+This at first did not work and I had to edit the `requirements.txt`. I removed <code>dataclasses==0.8</code> as this is no longer needed for Python 3.7 and higher. I opened a pull request back to the original Caxton repo since this was the only change needed for training to run.<br><br>
+
+<h1>Data</h1><br>
+
+5. I downloaded the data from <a href="https://www.repository.cam.ac.uk/items/6d77cd6d-8569-4bf4-9d5f-311ad2a49ac8">this dataset page</a>. Specifically, I downloaded <a href="https://www.repository.cam.ac.uk/bitstreams/54554b08-6ac5-495f-8d51-9c614d6839f7/download">caxton_dataset_filtered.csv</a>. I also consulted with Dr. Wilson and downloaded <a href="https://www.repository.cam.ac.uk/bitstreams/ec724f7c-14e3-464d-b031-372cd2765a03/download">print0.zip</a>, which I unzipped. I decided to train the model using only these images.<br><br>
+
+6. To train only with print0, I ran a script that drops entries not associated with the print0 dataset. This script is in <a href="https://github.com/philip-hub/caxton_fork/blob/main/extras/main.py">extras/main.py</a>. Make sure that the print0 images are placed in the appropriate path. See the README in the extras folder for more details.<br><br>
+
+7. I moved the `print0` folder to <code>data/caxton</code> and placed the outputted <code>caxton_data_filtered.csv</code> into the <code>data/</code> directory.<br><br>
+
+8. Final training step: start the training (ensure your virtual environment is activated):<br>
+<code>python src/train.py</code><br>
+This successfully trained my model!<br><br>
+
+<h1>Testing</h1>
+
+9. To run the test file, locate your checkpoint file after training. Mine was:<br>
+<code>src/checkpoints/21052025/1234/MHResAttNet-dataset_full-21052025-epoch=09-val_loss=37.10-val_acc=0.46.ckpt</code><br><br>
+
+Use Vim or a Visual Studio SSH client to open and edit `test.py`. On line 19, set the path:<br>
+<code>checkpoint_path = "/home/poundspb/Computer Vision/caxton_fork/src/checkpoints/21052025/1234/MHResAttNet-dataset_full-21052025-epoch=09-val_loss=37.10-val_acc=0.46.ckpt"</code><br><br>
+
+10. Run test.py:<br>
+<code>python src/test.py</code><br>
+This will show you the test accurcy and loss.
+
+<h1>Running with real world samples</h1>
+
+9. To run the sample file, locate your test image path. Upload your images to the path. Mine was:<br>
+<code>/home/poundspb/Computer Vision/caxton_fork/data/full/</code><br><br>
+
+Use Vim or a Visual Studio SSH client to open and edit `test.py`. On line 20, set the path you located in step 9:<br>
+<code>checkpoint_path = "/home/poundspb/Computer Vision/caxton_fork/src/checkpoints/21052025/1234/MHResAttNet-dataset_full-21052025-epoch=09-val_loss=37.10-val_acc=0.46.ckpt"</code> And then on line 29 change to your image path
+
+<code>sample_data = "/home/poundspb/Computer Vision/caxton_fork/data/full/"</code>
+
+Run samples.py
+<code>python src/samples.py</code><br>
+
+*Note I had to make a few changes to samples.py for it to work. 
+* Other note I used chatgpt on this project
+<br><br>
+

SAMPLE_OUTPUT.md

@@ -0,0 +1,46 @@
+Testing output
+<code>
+
+(env) poundspb@gebru:~/Computer Vision/caxton_fork/src$ python test.py
+Global seed set to 1234
+GPU available: True, used: True
+TPU available: None, using: 0 TPU cores
+LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1,2,3,4,5,6,7]
+Using native 16bit precision.
+Testing: 0it [00:00, ?it/s]/home/poundspb/Computer Vision/caxton_fork/env/lib/python3.9/site-packages/pytorch_lightning/core/step_result.py:145: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
+  value = torch.tensor(value, device=device, dtype=torch.float)
+Testing: 100%|██████████████████████████████████████████████████████████| 9/9 [00:01<00:00,  5.32it/s]
+--------------------------------------------------------------------------------
+DATALOADER:0 TEST RESULTS
+{'test_acc': tensor(0.5448, device='cuda:0'),
+ 'test_acc_epoch': tensor(0.5448, device='cuda:0'),
+ 'test_loss': tensor(3.7239, device='cuda:0'),
+ 'test_loss0': tensor(0.9406, device='cuda:0'),
+ 'test_loss1': tensor(0.9644, device='cuda:0'),
+ 'test_loss2': tensor(0.8445, device='cuda:0'),
+ 'test_loss3': tensor(0.9130, device='cuda:0'),
+ 'test_loss_epoch': tensor(3.6625, device='cuda:0')}
+</code>
+
+Sample Output
+
+<code>
+
+(env) poundspb@gebru:~/Computer Vision/caxton_fork/src$ python samples.py
+********* CAXTON sample predictions *********
+Flow rate | Lateral speed | Z offset | Hotend
+*********************************************
+Input: IMG_3438.jpg -> Prediction: [0 0 2 2]
+Input: IMG_3436.jpg -> Prediction: [0 0 2 2]
+Input: sample0021.jpg -> Prediction: [0 0 2 2]
+Input: sample1221.jpg -> Prediction: [0 0 2 2]
+Input: sample1111.jpg -> Prediction: [0 0 2 2]
+Input: IMG_3437.jpg -> Prediction: [0 0 2 2]
+Input: sample1011.jpg -> Prediction: [0 0 2 2]
+Input: sample2120.jpg -> Prediction: [0 0 2 2]
+Input: sample0022.jpg -> Prediction: [0 0 2 2]
+Input: sample2122.jpg -> Prediction: [0 0 2 2]
+Input: sample2011.jpg -> Prediction: [0 0 2 2]
+Completed 11 predictions in 1.84s
+
+</code>

data/caxton_dataset/readme.md

@@ -0,0 +1 @@
+After running the script move he unzipped files here