Rename README (2).md to README.md

README (2).md

@@ -1,44 +0,0 @@
----
-license: apache-2.0
-base_model: google/vit-base-patch16-224-in21k
-tags:
-- generated_from_trainer
-metrics:
-- accuracy
-model-index:
-- name: pose-estimation-front-side-back
-  results: []
----
-
-# Pose Estimation: front,side,back
-
-## Model description
-This model predicts the person's body position relative to the camera: **front, side, back**. It was trained on [Lucy in the Sky](https://www.lucyinthesky.com/shop) images.
-
-This model is a fine-tuned version of [google/vit-base-patch16-224-in21k](https://huggingface.co/google/vit-base-patch16-224-in21k).
-
-## Training and evaluation data
-It achieves the following results on the evaluation set:
-- Loss: 0.2524
-- Accuracy: 0.9355
-
-### Training hyperparameters
-
-The following hyperparameters were used during training:
-- learning_rate: 5e-05
-- train_batch_size: 16
-- eval_batch_size: 16
-- seed: 42
-- gradient_accumulation_steps: 4
-- total_train_batch_size: 64
-- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
-- lr_scheduler_type: linear
-- lr_scheduler_warmup_ratio: 0.1
-- num_epochs: 20
-
-### Framework versions
-
-- Transformers 4.34.0
-- Pytorch 2.0.1+cu118
-- Datasets 2.14.5
-- Tokenizers 0.14.0

README.md

@@ -0,0 +1,68 @@
+---
+license: apache-2.0
+base_model: swin_s3_base_224
+---
+
+# Feature Prediction
+
+## Model description
+
+
+## How to use
+
+```python
+import torch
+import torchvision.transforms as transforms
+from PIL import Image
+from safetensors.torch import load_model
+from huggingface_hub import hf_hub_download
+from timm import list_models, create_model
+import os
+import numpy as np
+
+# Download model from hub
+os.makedirs('/content/swin_s3_base_224', exist_ok=True)
+hf_hub_download(repo_id="LucyintheSky/lucy-feature-prediction", filename="model.safetensors", local_dir="/content/swin_s3_base_224")
+
+# Intialize the model
+model_name='swin_s3_base_224'
+model = create_model(
+    model_name,
+    num_classes=36
+)
+load_model(model,f'./{model_name}/model.safetensors')
+
+# Define class names
+class_names = ["3/4 Sleeve", "Accessory", "Babydoll", "Closed Back", "Corset", "Crochet", "Cutouts", "Draped", "Floral", "Gloves", "Halter", "Lace", "Long", "Long Sleeve", "Midi", "No Slit", "Off The Shoulder", "One Shoulder", "Open Back", "Pockets", "Print", "Puff Sleeve", "Ruched", "Satin", "Sequins", "Shimmer", "Short", "Short Sleeve", "Side Slit", "Square Neck", "Strapless", "Sweetheart Neck", "Tight", "V-Neck", "Velvet", "Wrap"]
+label2id = {c:idx for idx,c in enumerate(class_names)}
+id2label = {idx:c for idx,c in enumerate(class_names)}
+
+def predict_features(image_path):
+    # Load PIL image
+    pil_image = Image.open(image_path).convert('RGB')
+
+    # Define transformations to resize and convert image to tensor
+    transform = transforms.Compose([
+        transforms.Resize((224, 224)),
+        transforms.ToTensor()
+    ])
+    tensor_image = transform(pil_image)
+
+    inputs = tensor_image.unsqueeze(0)
+
+    with torch.no_grad():
+        logits = model(inputs)
+
+    # apply sigmoid activation to convert logits to probabilities
+    # getting labels with confidence threshold of 0.5
+    predictions = logits.sigmoid() > 0.5
+
+    # converting one-hot encoded predictions back to list of labels
+    predictions = predictions.float().numpy().flatten() # convert boolean predictions to float
+    pred_labels = np.where(predictions==1)[0] # find indices where prediction is 1
+    pred_labels = ([id2label[label] for label in pred_labels]) # converting integer labels to string
+
+    return pred_labels
+
+predict_features('image.jpg')
+```