Initial upload

main.py

@@ -0,0 +1,153 @@
+import datetime
+import os
+
+from torch.nn import Linear
+from torchvision.transforms import v2
+
+import data.dataset
+from torch.optim.lr_scheduler import CosineAnnealingLR
+import pandas
+from torchmetrics.classification import MulticlassAccuracy, MulticlassAveragePrecision, MulticlassF1Score
+from torchvision.models import mobilenet_v3_small, MobileNet_V3_Small_Weights
+from torch import nn, optim
+
+import torch
+from tqdm import tqdm
+
+from torch.utils.data import random_split
+
+import mlflow
+
+if __name__ == '__main__':
+    mlflow.set_tracking_uri('http://localhost:5000')
+    curr_date = datetime.datetime.now()
+    os.mkdir(f"outputs/{curr_date}")
+    # Input data files are available in the read-only "../input/" directory
+    # For example, running this (by clicking run or pressing Shift+Enter) will list all files under the input directory
+    with mlflow.start_run():
+        device = torch.device("cuda")
+        augmentation_transforms = v2.Compose([
+            v2.RandomHorizontalFlip(),
+            v2.RandomVerticalFlip(),
+            v2.RandomGrayscale(),
+            v2.RandomAutocontrast(),
+            v2.RandomRotation(45),
+        ]).to("cuda")
+        dataset = data.dataset.OrangeDataset("/home/jarric/orange_dataset/processed/FIELD IMAGES/")
+
+        train_size = int(0.75 * len(dataset))
+        val_size = len(dataset) - train_size
+        train_dataset, val_dataset = random_split(dataset, [train_size, val_size])
+
+        # Create data loaders
+        batch_size = 32
+
+        train_loader = torch.utils.data.DataLoader(
+            train_dataset,
+            batch_size=batch_size,
+            shuffle=False,
+            num_workers=8,
+
+        )
+
+        val_loader = torch.utils.data.DataLoader(
+            val_dataset,
+            batch_size=batch_size,
+            shuffle=False,
+            num_workers=8,
+
+        )
+        epochs = 100
+
+        mobilenet_v3_model = mobilenet_v3_small(weights=MobileNet_V3_Small_Weights.IMAGENET1K_V1.DEFAULT).to(device)
+
+        for param in mobilenet_v3_model.parameters(): # freeze layers
+            param.requires_grad = False
+
+        mobilenet_v3_model.classifier[3] = Linear(in_features=1024, out_features=3, bias=True)
+        mobilenet_v3_model.classifier[3].requires_grad = True
+
+        mobilenet_v3_model.cuda()
+        loss_fn = nn.CrossEntropyLoss().to(device)
+
+        optimizer = optim.AdamW(
+            mobilenet_v3_model.parameters(),
+            lr=1e-4,
+            weight_decay=1e-4)
+
+        reporting_interval_train = 50
+        reporting_interval_val = 10
+
+        loss_fn = loss_fn.to(device)
+
+        # metrics
+        acc_metric = MulticlassAccuracy(num_classes=3).to(device)
+        ap_metric = MulticlassAveragePrecision(num_classes=3, average="macro").to(device)
+        f1_metric = MulticlassF1Score(num_classes=3).to(device)
+
+
+        train_step = 0
+        val_step = 0
+        for epoch in tqdm(range(0, epochs)):
+            train_loss = 0
+            avg_accuracy = 0
+            cur_iter = 0
+            average_precision = 0
+            f1_score_avg = 0
+            mobilenet_v3_model.train()
+            for images, labels in tqdm(train_loader, leave=False):
+                images, labels = images.to(device), labels.to(device)
+
+                images = augmentation_transforms(images)
+                outputs = mobilenet_v3_model(images)
+                loss = loss_fn(outputs, labels)
+                train_loss += loss.item()
+
+                optimizer.zero_grad()
+                loss.backward()
+                optimizer.step()
+                _, predicted = torch.max(outputs.data, 1)
+                avg_accuracy += acc_metric(predicted, labels)
+                average_precision += ap_metric(outputs, labels)
+                f1_score_avg += f1_metric(predicted, labels)
+
+            train_loss /= len(train_loader)
+            avg_accuracy /= len(train_loader)
+            average_precision /= len(train_loader)
+            f1_score_avg /= len(train_loader)
+            mlflow.log_metric("train_loss", train_loss, step=epoch)
+            mlflow.log_metric("train_avg_accuracy", avg_accuracy, step=epoch)
+            mlflow.log_metric("train_average_precision", average_precision, step=epoch)
+            mlflow.log_metric("f1_score_avg", f1_score_avg, step=epoch)
+
+            val_loss = 0
+            val_accuracy = 0
+            cur_iter = 0
+            average_precision_val = 0
+            f1_score_avg_val = 0
+            mobilenet_v3_model.eval()
+            with torch.no_grad():
+                for images, labels in tqdm(val_loader, leave=False):
+                    images, labels = images.to(device), labels.to(device)
+
+                    # Forward pass
+                    outputs = mobilenet_v3_model(images)
+                    loss = loss_fn(outputs, labels)
+                    val_loss += loss.item()
+                    _, predicted = torch.max(outputs.data, 1)
+                    average_precision_val += ap_metric(outputs, labels)
+                    val_accuracy += acc_metric(predicted, labels)
+                    f1_score_avg_val += f1_metric(predicted, labels)
+
+            val_loss /= len(val_loader)
+            val_accuracy /= len(val_loader)
+            average_precision_val /= len(val_loader)
+            f1_score_avg_val /= len(val_loader)
+            mlflow.log_metric("val_loss", val_loss, step=epoch)
+            mlflow.log_metric("val_avg_accuracy", val_accuracy, step=epoch)
+            mlflow.log_metric("val_average_precision", average_precision_val, step=epoch)
+            mlflow.log_metric("val_f1_score_avg", f1_score_avg_val, step=epoch)
+
+
+            torch.save(mobilenet_v3_model, f"outputs/{curr_date}/model_{epoch}_finetuned.pt")
+            mlflow.log_artifact(f"outputs/{curr_date}/model_{epoch}_finetuned.pt")

model_99_finetuned.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e67254771b55408e9bb3310ff598ae747238fd218c3274d54875d07914c7426
+size 6255666

pretransform_files.py

@@ -0,0 +1,37 @@
+import os
+import pathlib
+from copy import copy
+from typing import List
+
+import torch
+from torchvision.io import read_image
+from torchvision.models import MobileNet_V3_Small_Weights
+from tqdm import tqdm
+
+if __name__ == "__main__":
+    target_types = ["*jpg", "*jpeg", "*png"]
+    target_dataset_path = pathlib.Path("/home/jarric/orange_dataset/archive/FIELD IMAGES/")
+    output_path = "/home/jarric/orange_dataset/processed/FIELD IMAGES/"
+    file_paths: List[pathlib.Path] = []
+
+    mobilenet_v3_transforms = MobileNet_V3_Small_Weights.IMAGENET1K_V1.transforms()
+    mobilenet_v3_transforms = mobilenet_v3_transforms.to("cuda")
+
+    for target_type in target_types:
+        file_paths.extend(list(target_dataset_path.rglob(target_type)))
+
+
+    for file_path in tqdm(file_paths, desc="Preprocessing files"):
+        image_tensor = read_image(str(file_path)).to("cuda")
+        processed_tensor = mobilenet_v3_transforms(image_tensor)
+
+        filename_parts = file_path.parts[-2:]
+        augmented_filename = f"{filename_parts[-1].split('.')[0]}_augmented.{filename_parts[-1].split('.')[-1]}"
+        augmented_file_parts = copy(list(filename_parts))
+        augmented_file_parts[-1] = augmented_filename
+        output_file_path = pathlib.Path(f"{output_path}{'/'.join(filename_parts)}")
+
+        if not output_file_path.parent.exists():
+            os.mkdir(str(output_file_path.parent))
+
+        torch.save(processed_tensor.cpu(), output_file_path)

read_cam.py

@@ -0,0 +1,83 @@
+import numpy as np
+import torch
+import cv2
+from sympy.stats.rv import probability
+from torch.nn import Sequential, Conv2d, BatchNorm2d, Hardswish, ReLU, AdaptiveAvgPool2d, Hardsigmoid, Linear, Dropout
+from torchvision.io import read_image
+from torchvision.models import MobileNet_V3_Small_Weights, MobileNetV3
+from torchvision.models.mobilenetv3 import InvertedResidual
+from torchvision.ops import Conv2dNormActivation, SqueezeExcitation
+
+if __name__ == "__main__":
+    torch.serialization.add_safe_globals([MobileNetV3])
+    torch.serialization.add_safe_globals([Sequential])
+    torch.serialization.add_safe_globals([Conv2dNormActivation])
+    torch.serialization.add_safe_globals([Conv2d])
+    torch.serialization.add_safe_globals([BatchNorm2d])
+    torch.serialization.add_safe_globals([Hardswish])
+    torch.serialization.add_safe_globals([InvertedResidual])
+    torch.serialization.add_safe_globals([ReLU])
+    torch.serialization.add_safe_globals([SqueezeExcitation])
+    torch.serialization.add_safe_globals([AdaptiveAvgPool2d])
+    torch.serialization.add_safe_globals([Hardsigmoid])
+    torch.serialization.add_safe_globals([Linear])
+    torch.serialization.add_safe_globals([Dropout])
+
+    mobilenetv3 = torch.load("/home/jarric/PycharmProjects/OrangeRecognizer/outputs/2025-04-21 12:08:02.020215/model_99_finetuned.pt")
+    mobilenetv3 = mobilenetv3.to("cuda")
+    mobilenet_v3_transforms = MobileNet_V3_Small_Weights.IMAGENET1K_V1.transforms().to("cuda")
+    class_assigment = {'citrus canker': 0, 'healthy': 1, 'melanose': 2}
+    class_numbering = {v:k for k, v in class_assigment.items()}
+
+    font = cv2.FONT_HERSHEY_SIMPLEX
+
+    vid_cam = cv2.VideoCapture(0)
+    confidence_list = []
+    if not vid_cam.isOpened():
+        print("Cannot open camera")
+        exit(-1)
+
+    warmed_up = False
+    final_confidence = 0
+    class_name = ""
+    averaged_final_confidences = []
+    while True:
+        ret, frame = vid_cam.read()
+        if not ret:
+             exit(-2)
+
+        orig_frame = frame.copy()
+        frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+        frame_resized_for_lookup = cv2.resize(orig_frame, (244 * 2, 224 * 2))
+        tensor_frame = torch.Tensor(frame).permute((2, 0, 1)).to(torch.uint8).to("cuda") # fuck you torch
+        processed_frame = mobilenet_v3_transforms(tensor_frame.unsqueeze(0))
+        logits = mobilenetv3(processed_frame)
+        confidences = logits.softmax(dim=-1)
+        _, classification = torch.max(logits.data, 1)
+        value, _ = torch.max(confidences, 1)
+        confidence_list.append(confidences.detach().cpu().squeeze(0))
+
+        if len(confidence_list) >= 50:
+            warmed_up = True
+
+            averaged_final_confidences = []
+            for class_num in class_numbering.keys():
+                averaged_final_confidences.append(np.array(confidence_list)[:, class_num].mean())
+
+            final_confidence, selected_class = torch.max(torch.Tensor(averaged_final_confidences).to("cuda").unsqueeze(0), 1)
+            class_name = class_numbering[selected_class.detach().cpu().item()]
+            averages_for_reference = [value.item() for value in averaged_final_confidences]
+            print(averages_for_reference)
+
+            confidence_list.clear()
+
+        if warmed_up:
+            cv2.putText(frame_resized_for_lookup, f"{class_name}, [{round(final_confidence.detach().cpu().item(), 2)}]", (50, 50), font, 1, (0, 0, 255), 3)
+            cv2.imshow("frame", frame_resized_for_lookup)
+
+        if cv2.waitKey(1) == ord('q'):
+            break
+
+    # vid_cam.release()
+    cv2.destroyAllWindows()
+    exit(0)

requirements.txt

@@ -0,0 +1,189 @@
+alembic==1.15.2
+annotated-types==0.7.0
+anyio @ file:///croot/anyio_1729121277521/work
+argon2-cffi @ file:///opt/conda/conda-bld/argon2-cffi_1645000214183/work
+argon2-cffi-bindings @ file:///croot/argon2-cffi-bindings_1736182440035/work
+asttokens @ file:///croot/asttokens_1743630435401/work
+async-lru @ file:///work/perseverance-python-buildout/croot/async-lru_1701732681408/work
+attrs @ file:///croot/attrs_1734533101012/work
+babel @ file:///croot/babel_1737454360933/work
+beautifulsoup4 @ file:///croot/beautifulsoup4-split_1718029820055/work
+bleach @ file:///croot/bleach_1732290411627/work
+blinker==1.9.0
+Bottleneck @ file:///croot/bottleneck_1731058641041/work
+Brotli @ file:///croot/brotli-split_1736182456865/work
+cachetools==5.5.2
+certifi @ file:///home/conda/feedstock_root/build_artifacts/certifi_1739515848642/work/certifi
+cffi @ file:///croot/cffi_1736182485317/work
+charset-normalizer @ file:///croot/charset-normalizer_1721748349566/work
+click==8.1.8
+cloudpickle==3.1.1
+comm @ file:///croot/comm_1709322850197/work
+contourpy @ file:///croot/contourpy_1732540045555/work
+cycler @ file:///tmp/build/80754af9/cycler_1637851556182/work
+databricks-sdk==0.50.0
+debugpy @ file:///croot/debugpy_1736267418885/work
+decorator @ file:///opt/conda/conda-bld/decorator_1643638310831/work
+defusedxml @ file:///tmp/build/80754af9/defusedxml_1615228127516/work
+Deprecated @ file:///home/conda/feedstock_root/build_artifacts/deprecated_1737986966356/work
+docker==7.1.0
+executing @ file:///opt/conda/conda-bld/executing_1646925071911/work
+fastapi==0.115.12
+fastjsonschema @ file:///croot/python-fastjsonschema_1731939362158/work
+filelock==3.13.1
+Flask==3.1.0
+fonttools @ file:///croot/fonttools_1737039080035/work
+fsspec==2024.6.1
+gitdb==4.0.12
+GitPython==3.1.44
+google-auth==2.39.0
+graphene==3.4.3
+graphql-core==3.2.6
+graphql-relay==3.2.0
+greenlet==3.2.0
+gunicorn==23.0.0
+h11 @ file:///croot/h11_1706652277403/work
+httpcore @ file:///croot/httpcore_1706728464539/work
+httpx @ file:///croot/httpx_1723474802858/work
+huggingface-hub==0.30.2
+idna @ file:///croot/idna_1714398848350/work
+importlib_metadata @ file:///home/conda/feedstock_root/build_artifacts/importlib-metadata_1737420181517/work
+inquirerpy==0.3.4
+ipykernel @ file:///croot/ipykernel_1737660677549/work
+ipython @ file:///croot/ipython_1734548052611/work
+ipywidgets==8.1.6
+itsdangerous==2.2.0
+jedi @ file:///croot/jedi_1733987392413/work
+Jinja2==3.1.4
+joblib==1.4.2
+json5 @ file:///croot/json5_1730786798687/work
+jsonschema @ file:///croot/jsonschema_1728486696720/work
+jsonschema-specifications @ file:///work/perseverance-python-buildout/croot/jsonschema-specifications_1701731664072/work
+jupyter-events @ file:///croot/jupyter_events_1741184577592/work
+jupyter-lsp @ file:///work/perseverance-python-buildout/croot/jupyter-lsp-meta_1707343107456/work
+jupyter_client @ file:///croot/jupyter_client_1737570961872/work
+jupyter_core @ file:///croot/jupyter_core_1718818295206/work
+jupyter_server @ file:///croot/jupyter_server_1741200034823/work
+jupyter_server_terminals @ file:///croot/jupyter_server_terminals_1744706698694/work
+jupyterlab @ file:///croot/jupyterlab_1737555423487/work
+jupyterlab_pygments @ file:///croot/jupyterlab_pygments_1741124142640/work
+jupyterlab_server @ file:///croot/jupyterlab_server_1725865349919/work
+jupyterlab_widgets==3.0.14
+kiwisolver @ file:///croot/kiwisolver_1737039087198/work
+lightning @ file:///home/conda/feedstock_root/build_artifacts/lightning_1742437252389/work
+lightning-utilities @ file:///home/conda/feedstock_root/build_artifacts/lightning-utilities_1743734367405/work
+lit @ file:///home/conda/feedstock_root/build_artifacts/llvm-package_1744799902975/work/llvm/utils/lit
+Mako==1.3.10
+Markdown==3.8
+MarkupSafe==2.1.5
+matplotlib==3.10.0
+matplotlib-inline @ file:///work/perseverance-python-buildout/croot/matplotlib-inline_1698864771271/work
+mistune @ file:///croot/mistune_1741124011532/work
+mkl-service==2.4.0
+mkl_fft @ file:///io/mkl313/mkl_fft_1730824109137/work
+mkl_random @ file:///io/mkl313/mkl_random_1730823916628/work
+mlflow==2.21.3
+mlflow-skinny==2.21.3
+mpmath @ file:///home/conda/feedstock_root/build_artifacts/mpmath_1733302684489/work
+nbclient @ file:///croot/nbclient_1741123995822/work
+nbconvert @ file:///croot/nbconvert-meta_1741184653315/work
+nbformat @ file:///croot/nbformat_1728049424075/work
+nest-asyncio @ file:///croot/nest-asyncio_1708532673751/work
+networkx==3.3
+notebook @ file:///croot/notebook_1738159946465/work
+notebook_shim @ file:///croot/notebook-shim_1741707758683/work
+numexpr @ file:///croot/numexpr_1730215937391/work
+numpy==2.1.2
+nvidia-cublas-cu12==12.4.5.8
+nvidia-cuda-cupti-cu12==12.4.127
+nvidia-cuda-nvrtc-cu12==12.4.127
+nvidia-cuda-runtime-cu12==12.4.127
+nvidia-cudnn-cu12==9.1.0.70
+nvidia-cufft-cu12==11.2.1.3
+nvidia-curand-cu12==10.3.5.147
+nvidia-cusolver-cu12==11.6.1.9
+nvidia-cusparse-cu12==12.3.1.170
+nvidia-cusparselt-cu12==0.6.2
+nvidia-nccl-cu12==2.21.5
+nvidia-nvjitlink-cu12==12.4.127
+nvidia-nvtx-cu12==12.4.127
+opencv-python==4.11.0.86
+opentelemetry-api @ file:///home/conda/feedstock_root/build_artifacts/opentelemetry-api_1744805275188/work
+opentelemetry-sdk==1.32.1
+opentelemetry-semantic-conventions==0.53b1
+overrides @ file:///work/perseverance-python-buildout/croot/overrides_1701732220415/work
+packaging @ file:///croot/packaging_1734472117206/work
+pandas @ file:///croot/pandas_1732735089971/work/dist/pandas-2.2.3-cp312-cp312-linux_x86_64.whl#sha256=57b66702d418720ec8483f7c4ec7c08d41815316ad7ce09d5b7bbc34eefcfdfd
+pandocfilters @ file:///opt/conda/conda-bld/pandocfilters_1643405455980/work
+parso @ file:///croot/parso_1733963305961/work
+pexpect @ file:///tmp/build/80754af9/pexpect_1605563209008/work
+pfzy==0.3.4
+pillow==11.0.0
+Pillow-SIMD==9.5.0.post2
+platformdirs @ file:///croot/platformdirs_1744273042065/work
+prometheus_client @ file:///croot/prometheus_client_1744271615306/work
+prompt-toolkit @ file:///croot/prompt-toolkit_1704404351921/work
+protobuf==5.29.4
+psutil @ file:///croot/psutil_1736367091698/work
+ptyprocess @ file:///tmp/build/80754af9/ptyprocess_1609355006118/work/dist/ptyprocess-0.7.0-py2.py3-none-any.whl
+pure-eval @ file:///opt/conda/conda-bld/pure_eval_1646925070566/work
+pyarrow==19.0.1
+pyasn1==0.6.1
+pyasn1_modules==0.4.2
+pycparser @ file:///tmp/build/80754af9/pycparser_1636541352034/work
+pydantic==2.11.3
+pydantic_core==2.33.1
+Pygments @ file:///croot/pygments_1744664109463/work
+pyparsing @ file:///croot/pyparsing_1731445506121/work
+PySocks @ file:///work/perseverance-python-buildout/croot/pysocks_1698845478203/work
+python-dateutil @ file:///croot/python-dateutil_1716495738603/work
+python-json-logger @ file:///croot/python-json-logger_1734370021104/work
+pytorch-lightning @ file:///home/conda/feedstock_root/build_artifacts/pytorch-lightning_1742579345300/work
+pytz @ file:///croot/pytz_1713974312559/work
+PyYAML @ file:///croot/pyyaml_1728657952215/work
+pyzmq @ file:///croot/pyzmq_1734687138743/work
+referencing @ file:///work/perseverance-python-buildout/croot/referencing_1701731622327/work
+requests @ file:///croot/requests_1730999120400/work
+rfc3339-validator @ file:///work/perseverance-python-buildout/croot/rfc3339-validator_1698873973408/work
+rfc3986-validator @ file:///work/perseverance-python-buildout/croot/rfc3986-validator_1698874010100/work
+rpds-py @ file:///croot/rpds-py_1736541261634/work
+rsa==4.9.1
+scikit-learn==1.6.1
+scipy==1.15.2
+seaborn @ file:///croot/seaborn_1741185878286/work
+Send2Trash @ file:///croot/send2trash_1736540790175/work
+setuptools==75.8.0
+six @ file:///croot/six_1744271502820/work
+smmap==5.0.2
+sniffio @ file:///croot/sniffio_1705431295498/work
+soupsieve @ file:///work/perseverance-python-buildout/croot/soupsieve_1698866207280/work
+SQLAlchemy==2.0.40
+sqlparse==0.5.3
+stack-data @ file:///opt/conda/conda-bld/stack_data_1646927590127/work
+starlette==0.46.2
+sympy==1.13.1
+terminado @ file:///work/perseverance-python-buildout/croot/terminado_1698874934826/work
+threadpoolctl==3.6.0
+tinycss2 @ file:///croot/tinycss2_1738337643607/work
+torch==2.6.0
+torchaudio==2.6.0
+torchmetrics @ file:///home/conda/feedstock_root/build_artifacts/torchmetrics_1744107217838/work
+torchvision==0.21.0
+tornado @ file:///croot/tornado_1733960490606/work
+tqdm @ file:///croot/tqdm_1738943501192/work
+traitlets @ file:///croot/traitlets_1718227057033/work
+triton==3.2.0
+typing-inspection==0.4.0
+typing_extensions @ file:///croot/typing_extensions_1734714854207/work
+tzdata @ file:///croot/python-tzdata_1690578112552/work
+unicodedata2 @ file:///croot/unicodedata2_1736541023050/work
+urllib3 @ file:///croot/urllib3_1737133630106/work
+uvicorn==0.34.2
+wcwidth @ file:///Users/ktietz/demo/mc3/conda-bld/wcwidth_1629357192024/work
+webencodings @ file:///work/perseverance-python-buildout/croot/webencodings_1698866454420/work
+websocket-client @ file:///croot/websocket-client_1715878298792/work
+Werkzeug==3.1.3
+wheel==0.45.1
+widgetsnbextension==4.0.14
+wrapt @ file:///croot/wrapt_1736540904746/work
+zipp @ file:///home/conda/feedstock_root/build_artifacts/zipp_1732827521216/work