Telegram Stickers Classification Model
This repository contains a pre-trained image classification model based on the YOLOv8m-cls for classifying Telegram stickers.
Model Details
- Model Size: 128x128 pixels
- Number of Classes: 1276
The training set contained 605,043 sticker images, each labeled with the Unicode code assigned to it based on the emoji representation provided by the author. For example, the Unicode U+1F917 represents the 🤗 emoji.
The dataset was created by extracting stickers from a total of 23,681 sets of stickers in Telegram. Sets that had only one emoji assigned to all stickers were not included in the dataset. This ensures a diverse range of stickers with different visual characteristics.
- Example images:
- U+1F604 0.12, U+1F606 0.10, U+1F602 0.07, U+1F601 0.06, U+1F603 0.04 (😄 0.12, 😆 0.10, 😂 0.07, 😁 0.06, 😃 0.04)
- U+1F52B 0.61, U+1F621 0.02, U+1F31F 0.02, U+1F497 0.01, U+1F620 0.01 (🔫 0.61, 😡 0.02, 🌟 0.02, 💗 0.01, 😠 0.01)
- U+1F610 0.25, U+1F642 0.23, U+1F431 0.05, U+1F60A 0.04, U+1F633 0.04 (😐 0.25, 🙂 0.23, 🐱 0.05, 😊 0.04, 😳 0.04)
- U+1F601 0.29, U+1F604 0.09, U+1F605 0.08, U+270C 0.05, U+1F33B 0.03 (😁 0.29, 😄 0.09, 😅 0.08, ✌ 0.05, 🌻 0.03)
- U+1F62D 0.34, U+1F622 0.20, U+1F97A 0.09, U+1F5A4 0.04, U+1F614 0.03 (😭 0.34, 😢 0.20, 🥺 0.09, 🖤 0.04, 😔 0.03)