Update README.md

README.md

@@ -1777,16 +1777,15 @@ configs:
 license: mit
 task_categories:
 - image-classification
+- video-classification
 tags:
 - neuromorphic
 - snn
-- spiking-neural-networks
-- event-based-vision
+- spiking neural networks
+- event
 - dvs
-
 - biology
 - pytorch
-
 - imagenet
 - cifar10
 - cifar100
@@ -1796,6 +1795,8 @@ tags:
 - fashionmnist
 - mini-imagenet
 pretty_name: I2E Neuromorphic Dataset
+language:
+- en
 ---
 
 
@@ -1814,62 +1815,167 @@ pretty_name: I2E Neuromorphic Dataset
 
 ## 🚀 Introduction
 
-This repository contains the **I2E-Datasets** for the paper **"I2E: Real-Time Image-to-Event Conversion for High-Performance Spiking Neural Networks"**, which has been accepted for **Oral Presentation at AAAI 2026**.
+This repository hosts the **I2E-Datasets**, a comprehensive suite of neuromorphic datasets generated using the **I2E (Image-to-Event)** framework. This work has been accepted for **Oral Presentation at AAAI 2026**.
 
-**I2E** is a pioneering framework that bridges the data scarcity gap in neuromorphic computing. By simulating microsaccadic eye movements via highly parallelized convolution, I2E converts static images into high-fidelity event streams in real-time (>300x faster than prior methods).
+**I2E** bridges the data scarcity gap in **Neuromorphic Computing** and **Spiking Neural Networks (SNNs)**. By simulating microsaccadic eye movements via highly parallelized convolution, I2E converts static images into high-fidelity event streams in real-time (>300x faster than prior methods).
 
-### ✨ Key Highlights
-* **SOTA Performance**: Achieves **60.50%** top-1 accuracy on Event-based ImageNet.
-* **Sim-to-Real Transfer**: Pre-training on I2E data enables **92.5%** accuracy on real-world CIFAR10-DVS, setting a new benchmark.
-* **Real-Time Conversion**: Enables on-the-fly data augmentation for deep SNN training.
+## 👁️ Visualization
 
+The following comparisons illustrate the high-fidelity conversion from static RGB images to dynamic event streams using I2E.
 
-## 💻 Usage
-The generated I2E-Datasets are provided in the [**Files and versions**](https://huggingface.co/datasets/UESTC-BICS/I2E/tree/main) section.
+<table border="0" style="width: 100%">
+  <tr>
+    <td width="25%" align="center"><img src="./assets/original_1.jpg" alt="Original 1" style="width:100%"></td>
+    <td width="25%" align="center"><img src="./assets/converted_1.gif" alt="Converted 1" style="width:100%"></td>
+    <td width="25%" align="center"><img src="./assets/original_2.jpg" alt="Original 2" style="width:100%"></td>
+    <td width="25%" align="center"><img src="./assets/converted_2.gif" alt="Converted 2" style="width:100%"></td>
+  </tr>
+  <tr>
+    <td width="25%" align="center"><img src="./assets/original_3.jpg" alt="Original 3" style="width:100%"></td>
+    <td width="25%" align="center"><img src="./assets/converted_3.gif" alt="Converted 3" style="width:100%"></td>
+    <td width="25%" align="center"><img src="./assets/original_4.jpg" alt="Original 4" style="width:100%"></td>
+    <td width="25%" align="center"><img src="./assets/converted_4.gif" alt="Converted 4" style="width:100%"></td>
+  </tr>
+</table>
+
+*More visualization comparisons can be found in [Visualization.md](./Visualization.md).*
+
+## 📦 Dataset Catalog
+
+We provide a comprehensive collection of standard benchmarks converted into event streams via the I2E algorithm.
+
+### 1. Standard Benchmarks (Classification)
+| Config Name | Original Source | Resolution $(H, W)$ | I2E Ratio | Event Rate | Samples (Train/Val) |
+| :--- | :--- | :--- | :--- | :--- | :--- |
+| **`I2E-CIFAR10`** | CIFAR-10 | 128 x 128 | 0.07 | 5.86% | 50k / 10k |
+| **`I2E-CIFAR100`** | CIFAR-100 | 128 x 128 | 0.07 | 5.76% | 50k / 10k |
+| **`I2E-ImageNet`** | ILSVRC2012 | 224 x 224 | 0.12 | 6.66% | 1.28M / 50k |
 
-### 🚀 Quick Start (Copy & Paste)
+### 2. Transfer Learning & Fine-grained
+| Config Name | Original Source | Resolution $(H, W)$ | I2E Ratio | Event Rate | Samples |
+| :--- | :--- | :--- | :--- | :--- | :--- |
+| **`I2E-Caltech101`** | Caltech-101 | 224 x 224 | 0.12 | 6.25% | 8.677k |
+| **`I2E-Caltech256`** | Caltech-256 | 224 x 224 | 0.12 | 6.04% | 30.607k |
+| **`I2E-Mini-ImageNet`**| Mini-ImageNet | 224 x 224 | 0.12 | 6.65% | 60k |
 
-You don't need to manually handle the bit-unpacking. Just use the following snippet to load the data directly into PyTorch tensors.
+### 3. Small Scale / Toy
+| Config Name | Original Source | Resolution $(H, W)$ | I2E Ratio | Event Rate | Samples |
+| :--- | :--- | :--- | :--- | :--- | :--- |
+| **`I2E-MNIST`** | MNIST | 32 x 32 | 0.10 | 9.56% | 60k / 10k |
+| **`I2E-FashionMNIST`** | Fashion-MNIST | 32 x 32 | 0.15 | 10.76% | 60k / 10k |
+
+> 🔜 **Coming Soon:** Object Detection and Semantic Segmentation datasets.
+
+## 🛠️ Preprocessing Protocol
+
+To ensure reproducibility, we specify the exact data augmentation pipeline applied to the static images **before** I2E conversion. 
+
+The `(H, W)` in the code below corresponds to the "Resolution" column in the Dataset Catalog above.
 
 ```python
-import os
-import torch
-from tqdm import tqdm
-from load_i2e_datasets import I2E_Dataset
-os.environ['HF_ENDPOINT'] = 'https://hf-mirror.com'            # Use HF mirror server in some regions
+from torchvision.transforms import v2
 
-model_cache_path = 'Your path to model cache'
-load_datasets = 'I2E-CIFAR10'
+# Standard Pre-processing Pipeline used for I2E generation
+transform_train = v2.Compose([
+    # Ensure 3-channel RGB (crucial for grayscale datasets like MNIST)
+    v2.Lambda(lambda x: x.convert('RGB')),
+    v2.PILToTensor(),
+    v2.Resize((H, W), interpolation=v2.InterpolationMode.BICUBIC),
+    v2.ToDtype(torch.float32, scale=True),
+])
+````
 
-train_dataset = I2E_Dataset(model_cache_path, load_datasets, split='train', transform=None, target_transform=None)
-val_dataset = I2E_Dataset(model_cache_path, load_datasets, split='validation', transform=None, target_transform=None)
-print(f"Train samples: {len(train_dataset)}, Validation samples: {len(val_dataset)}")
+## 💻 Usage
 
-# You can create dataloaders according to your needs
-train_loader = torch.utils.data.DataLoader(train_dataset, batch_size=256, shuffle=False, num_workers=32, pin_memory=False)
-val_loader = torch.utils.data.DataLoader(val_dataset, batch_size=256, shuffle=False, num_workers=32, pin_memory=False)
-```
+### 🚀 Quick Start
 
+You **do not** need to download any extra scripts. Just copy the code below. It handles the binary unpacking (converting Parquet bytes to PyTorch Tensors) automatically.
 
+```python
+import io
+import torch
+import numpy as np
+from datasets import load_dataset
+from torch.utils.data import Dataset, DataLoader
 
-This repository hosts the **datasets only**.
+# ==================================================================
+# 1. Core Decoding Function (Handles the binary packing)
+# ==================================================================
+def unpack_event_data(item, use_io=True):
+    """
+    Decodes the custom binary format:
+    Header (8 bytes) -> Shape (T, C, H, W) -> Body (Packed Bits)
+    """
+    if use_io:
+        with io.BytesIO(item['data']) as f:
+            raw_data = np.load(f)
+    else:
+        raw_data = np.load(item)
+        
+    header_size = 4 * 2      # Parse Header (First 8 bytes for 4 uint16 shape values)
+    shape_header = raw_data[:header_size].view(np.uint16)
+    original_shape = tuple(shape_header) # Returns (T, C, H, W)
+    
+    packed_body = raw_data[header_size:]    # Parse Body & Bit-unpacking
+    unpacked = np.unpackbits(packed_body)
+    
+    num_elements = np.prod(original_shape)  # Extract valid bits (Handle padding)
+    event_flat = unpacked[:num_elements]
+    event_data = event_flat.reshape(original_shape).astype(np.float32).copy()
+    
+    return torch.from_numpy(event_data)
 
-For the **I2E dataset generation code**, **training scripts**, and detailed usage instructions, please refer to our official GitHub repository.
+# ==================================================================
+# 2. Dataset Wrapper
+# ==================================================================
+class I2E_Dataset(Dataset):
+    def __init__(self, cache_dir, config_name, split='train', transform=None, target_transform=None):
+        print(f"🚀 Loading {config_name} [{split}] from Hugging Face...")
+        self.ds = load_dataset('UESTC-BICS/I2E', config_name, split=split, cache_dir=cache_dir, keep_in_memory=False)
+        self.transform = transform
+        self.target_transform = target_transform
 
-To generate the I2E-Datasets yourself using the I2E algorithm, please follow the instructions in the GitHub README.
+    def __len__(self):
+        return len(self.ds)
 
-[![GitHub](https://img.shields.io/badge/GitHub-Repository-black?logo=github)](https://github.com/Ruichen0424/I2E)
+    def __getitem__(self, idx):
+        item = self.ds[idx]
+        event = unpack_event_data(item)
+        label = item['label']
+        if self.transform:
+            event = self.transform(event)
+        if self.target_transform:
+            label = self.target_transform(label)
+        return event, label
 
+# ==================================================================
+# 3. Run Example
+# ==================================================================
+if __name__ == "__main__":
+    import os
+    os.environ['HF_ENDPOINT'] = 'https://hf-mirror.com'     # Use HF mirror server in some regions
 
-## 🏆 Model Zoo & Results
+    DATASET_NAME = 'I2E-CIFAR10'                            # Choose your config: 'I2E-CIFAR10', 'I2E-ImageNet', etc.
+    MODEL_PATH = 'Your cache path here'                     # e.g., './hf_datasets_cache/'
+    
+    train_dataset = I2E_Dataset(MODEL_PATH, DATASET_NAME, split='train')
+    val_dataset = I2E_Dataset(MODEL_PATH, DATASET_NAME, split='validation')
 
-We provide pre-trained models for **I2E-CIFAR** and **I2E-ImageNet**. You can download the `.pth` files directly from the [**Files and versions**](https://huggingface.co/Ruichen0424/I2E/tree/main) tab in model repository.
+    train_loader = DataLoader(train_dataset, batch_size=32, shuffle=True, num_workers=32, persistent_workers=True)
+    val_loader = DataLoader(val_dataset, batch_size=32, shuffle=False, num_workers=32, persistent_workers=True)
 
-[![Hugging Face](https://img.shields.io/badge/Hugging%20Face-Models-FFD21E?style=flat-square&logo=huggingface&logoColor=black)](https://huggingface.co/Ruichen0424/I2E)
+    events, labels = next(iter(train_loader))
+    print(f"✅ Loaded Batch Shape: {events.shape}") # Expect: [32, T, 2, H, W]
+    print(f"✅ Labels: {labels}")
+```
+
+## 🏆 Results (SOTA)
+
+Our I2E-pretraining sets new benchmarks for Sim-to-Real transfer on **CIFAR10-DVS**.
 
 <table border="1">
     <tr>
-        <th>Target Dataset</th>
+        <th>Dataset</th>
         <th align="center">Architecture</th>
         <th align="center">Method</th>
         <th align="center">Top-1 Acc</th>
@@ -1891,90 +1997,12 @@ We provide pre-trained models for **I2E-CIFAR** and **I2E-ImageNet**. You can do
         <td align="center" style="vertical-align: middle;">Transfer-II (Sim-to-Real)</td>
         <td align="center" style="vertical-align: middle;"><strong>92.5%</strong></td>
     </tr>
-    <!-- I2E-CIFAR10 -->
-    <tr>
-        <td rowspan="3" align="center" style="vertical-align: middle;"><strong>I2E-CIFAR10</strong></td>
-        <td align="center" style="vertical-align: middle;">MS-ResNet18</td>
-        <td align="center" style="vertical-align: middle;">Baseline-I</td>
-        <td align="center" style="vertical-align: middle;">85.07%</td>
-    </tr>
-    <tr>
-        <td align="center" style="vertical-align: middle;">MS-ResNet18</td>
-        <td align="center" style="vertical-align: middle;">Baseline-II</td>
-        <td align="center" style="vertical-align: middle;">89.23%</td>
-    </tr>
-    <tr>
-        <td align="center" style="vertical-align: middle;">MS-ResNet18</td>
-        <td align="center" style="vertical-align: middle;">Transfer-I</td>
-        <td align="center" style="vertical-align: middle;"><strong>90.86%</strong></td>
-    </tr>
-    <!-- I2E-CIFAR100 -->
-    <tr>
-        <td rowspan="3" align="center" style="vertical-align: middle;"><strong>I2E-CIFAR100</strong></td>
-        <td align="center" style="vertical-align: middle;">MS-ResNet18</td>
-        <td align="center" style="vertical-align: middle;">Baseline-I</td>
-        <td align="center" style="vertical-align: middle;">51.32%</td>
-    </tr>
-    <tr>
-        <td align="center" style="vertical-align: middle;">MS-ResNet18</td>
-        <td align="center" style="vertical-align: middle;">Baseline-II</td>
-        <td align="center" style="vertical-align: middle;">60.68%</td>
-    </tr>
-    <tr>
-        <td align="center" style="vertical-align: middle;">MS-ResNet18</td>
-        <td align="center" style="vertical-align: middle;">Transfer-I</td>
-        <td align="center" style="vertical-align: middle;"><strong>64.53%</strong></td>
-    </tr>
-    <!-- I2E-ImageNet -->
-    <tr>
-        <td rowspan="4" align="center" style="vertical-align: middle;"><strong>I2E-ImageNet</strong></td>
-        <td align="center" style="vertical-align: middle;">MS-ResNet18</td>
-        <td align="center" style="vertical-align: middle;">Baseline-I</td>
-        <td align="center" style="vertical-align: middle;">48.30%</td>
-    </tr>
-    <tr>
-        <td align="center" style="vertical-align: middle;">MS-ResNet18</td>
-        <td align="center" style="vertical-align: middle;">Baseline-II</td>
-        <td align="center" style="vertical-align: middle;">57.97%</td>
-    </tr>
-    <tr>
-        <td align="center" style="vertical-align: middle;">MS-ResNet18</td>
-        <td align="center" style="vertical-align: middle;">Transfer-I</td>
-        <td align="center" style="vertical-align: middle;">59.28%</td>
-    </tr>
-    <tr>
-        <td align="center" style="vertical-align: middle;">MS-ResNet34</td>
-        <td align="center" style="vertical-align: middle;">Baseline-II</td>
-        <td align="center" style="vertical-align: middle;"><strong>60.50%</strong></td>
-    </tr>
 </table>
 
-> **Method Legend:**
-> * **Baseline-I**: Training from scratch with minimal augmentation.
-> * **Baseline-II**: Training from scratch with full augmentation.
-> * **Transfer-I**: Fine-tuning from Static ImageNet (or I2E-ImageNet for CIFAR targets).
-> * **Transfer-II**: Fine-tuning from I2E-CIFAR10.
-
-## 👁️ Visualization
-
-Below is the visualization of the I2E conversion process. We illustrate the high-fidelity conversion from static RGB images to dynamic event streams.
-
-More than 200 additional visualization comparisons can be found in [Visualization.md](./Visualization.md).
+*For full results and model weights, please visit our [GitHub Repo](https://github.com/Ruichen0424/I2E).*
 
-<table border="0" style="width: 100%">
-  <tr>
-    <td width="25%" align="center"><img src="./assets/original_1.jpg" alt="Original 1" style="width:100%"></td>
-    <td width="25%" align="center"><img src="./assets/converted_1.gif" alt="Converted 1" style="width:100%"></td>
-    <td width="25%" align="center"><img src="./assets/original_2.jpg" alt="Original 2" style="width:100%"></td>
-    <td width="25%" align="center"><img src="./assets/converted_2.gif" alt="Converted 2" style="width:100%"></td>
-  </tr>
-  <tr>
-    <td width="25%" align="center"><img src="./assets/original_3.jpg" alt="Original 3" style="width:100%"></td>
-    <td width="25%" align="center"><img src="./assets/converted_3.gif" alt="Converted 3" style="width:100%"></td>
-    <td width="25%" align="center"><img src="./assets/original_4.jpg" alt="Original 4" style="width:100%"></td>
-    <td width="25%" align="center"><img src="./assets/converted_4.gif" alt="Converted 4" style="width:100%"></td>
-  </tr>
-</table>
+[![GitHub](https://img.shields.io/badge/GitHub-Repository-black?logo=github)](https://github.com/Ruichen0424/I2E)
+[![Hugging Face](https://img.shields.io/badge/Hugging%20Face-Models-FFD21E?style=flat-square&logo=huggingface&logoColor=black)](https://huggingface.co/Ruichen0424/I2E)
 
 ## 📜 Citation