Upload InkDetectionPipeline

README.md

@@ -0,0 +1,199 @@
+---
+library_name: transformers
+tags: []
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]

config.json

@@ -0,0 +1,28 @@
+{
+  "_name_or_path": "scrollprize/timesformer_GP_scroll1",
+  "architectures": [
+    "TimesformerScrollprizeModel"
+  ],
+  "auto_map": {
+    "AutoConfig": "scrollprize/timesformer_GP_scroll1--timesformer_config.TimesformerScrollprizeConfig",
+    "AutoModel": "scrollprize/timesformer_GP_scroll1--timesformer_model.TimesformerScrollprizeModel"
+  },
+  "custom_pipelines": {
+    "ink-detection": {
+      "impl": "ink_detection_pipeline.InkDetectionPipeline",
+      "pt": [
+        "AutoModel"
+      ],
+      "tf": []
+    }
+  },
+  "depth": 8,
+  "dim": 512,
+  "n_heads": 6,
+  "num_classes": 16,
+  "num_frames": 26,
+  "patch_size": 16,
+  "torch_dtype": "float32",
+  "transformers_version": "4.46.3",
+  "window_size": 64
+}

ink_detection_pipeline.py

@@ -0,0 +1,138 @@
+import torch
+import torch.nn.functional as F
+import numpy as np
+from transformers import Pipeline,AutoModel
+from tqdm import tqdm
+
+
+
+class InkDetectionPipeline(Pipeline):
+    """
+    A custom pipeline that:
+    1. Takes in a 3D image: shape (m, n, d).
+    2. Cuts it into (64, 64, d) tiles with a given stride.
+    3. Runs the model inference on each tile (model is 3D-to-2D).
+    4. Reconstructs the predictions into a full-size output.
+    """
+
+    def __init__(self, model, device='cuda', tile_size=64, stride=32, scale_factor=16,batch_size=32,**kwargs):
+        super().__init__(model=model, tokenizer=None, device=0 if device=='cuda' else -1)
+        self.model = model.to(device)
+        self.device = device
+        self.tile_size = tile_size
+        self.stride = stride
+        self.scale_factor = scale_factor
+        self.batch_size=batch_size
+    def preprocess(self, inputs):
+        """
+        inputs: np.ndarray of shape (m, n, d)
+        This function cuts the input volume into tiles of shape (tile_size, tile_size, d)
+        with a given stride.
+        Returns:
+          tiles: list of np arrays each (tile_size, tile_size, d)
+          coords: list of (x1, y1, x2, y2) coords
+        """
+        volume = inputs
+        m, n, d = volume.shape
+        tiles = []
+        coords = []
+        # Extract patches with overlap
+        for y in range(0, m - self.tile_size + 1, self.stride):
+            for x in range(0, n - self.tile_size + 1, self.stride):
+                y1, y2 = y, y + self.tile_size
+                x1, x2 = x, x + self.tile_size
+                patch = volume[y1:y2, x1:x2]  # shape (64,64,d)
+                tiles.append(patch.transpose(2,0,1))
+                coords.append((x1, y1, x2, y2))
+        return np.array(tiles,dtype=np.float16), coords, (m, n)
+    def _forward(self, model_inputs):
+        """
+        model_inputs: a list of patches (B, tile_size, tile_size, d)
+        The model expects input: (B, C=1, H=tile_size, W=tile_size)
+        and returns (B, 1, H=tile_size, W=tile_size).
+
+        We'll add batching using a for loop. We assume `self.batch_size` is defined.
+        """
+
+        patches = model_inputs
+        B = len(patches)
+        all_preds = []
+        # Process in batches to save memory
+        for start_idx in tqdm(range(0, B, self.batch_size)):
+            end_idx = start_idx + self.batch_size
+            sub_batch = torch.from_numpy(patches[start_idx:end_idx].astype(np.float32))  # shape: (subB, d, tile_size, tile_size)
+
+            # Add channel dimension: (subB, 1, tile_size, tile_size)
+            sub_batch = sub_batch.unsqueeze(1)
+
+            with torch.no_grad(), torch.autocast(self.device if self.device == 'cuda' else 'cpu'):
+                sub_y_preds = self.model(sub_batch.to(self.device))  # (subB, 1, tile_size, tile_size)
+
+            # Apply sigmoid
+            sub_y_preds = torch.sigmoid(sub_y_preds)
+
+            # Move to CPU and numpy
+            sub_y_preds = sub_y_preds.detach().cpu().numpy()
+            # shape (subB, 1, tile_size, tile_size)
+
+            all_preds.append(sub_y_preds)
+
+        # Concatenate along the batch dimension
+        y_preds = np.concatenate(all_preds, axis=0)  # (B, 1, tile_size, tile_size)
+
+        return y_preds
+
+    def postprocess(self, model_outputs, coords, full_shape):
+        """
+        model_outputs: np.ndarray of shape (B, 1, tile_size, tile_size)
+        coords: list of (x1, y1, x2, y2) for each tile
+        full_shape: (m,n)
+
+        We need to:
+        - Place each tile prediction into a full (m,n) array
+        - Use the kernel to weight and sum predictions
+        - Divide by count
+        - Optionally upsample by scale_factor if required
+        """
+        m, n = full_shape
+        # We will create mask_pred and mask_count to accumulate predictions
+        mask_pred = np.zeros((m, n), dtype=np.float32)
+        mask_count = np.zeros((m, n), dtype=np.float32)
+        B = model_outputs.shape[0]
+        # Interpolate (upsample) each prediction if needed
+        # Using PyTorch interpolate:
+        preds_tensor = torch.from_numpy(model_outputs.astype(np.float32))  # (B,1,64,64)
+        if self.scale_factor != 1:
+            preds_tensor = F.interpolate(
+                preds_tensor, scale_factor=self.scale_factor, mode='bilinear', align_corners=False
+            )
+        preds_tensor = preds_tensor.squeeze(1).numpy()  # (B, H_out, W_out)
+
+        out_tile_size = self.tile_size 
+
+        for i, (x1, y1, x2, y2) in enumerate(coords):
+            # Adjust coords due to upsampling
+            y2_up = y1 + out_tile_size
+            x2_up = x1 + out_tile_size
+
+            mask_pred[y1:y2_up, x1:x2_up] += preds_tensor[i] 
+            mask_count[y1:y2_up, x1:x2_up] += np.ones((out_tile_size, out_tile_size), dtype=np.float32)
+
+        mask_pred = np.divide(mask_pred, mask_count, out=np.zeros_like(mask_pred), where=mask_count!=0)
+        
+        return mask_pred
+    def _sanitize_parameters(self,**kwargs):
+        return {},{},{}
+    def __call__(self, image: np.ndarray):
+        """
+        Args:
+            image: np.ndarray of shape (m, n, d) input volume.
+        Returns:
+            mask_pred: np.ndarray of shape (m_out, n_out) predicted mask.
+        """
+        tiles, coords, full_shape = self.preprocess(image)
+        # Process in batches if too large (optional). Here we do a single batch inference for simplicity.
+        # If large images, consider chunking tiles into smaller batches.
+        outputs = self._forward(tiles)
+        mask_pred = self.postprocess(outputs, coords, full_shape)
+        return mask_pred

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:490a98f9491e1180274ed3a0c0a9c611d73a0109c0e0c0fbba1097562a972488
+size 151853128