C

Convnextv2 Atto.fcmae

Developed by timm
A self-supervised feature representation model based on ConvNeXt-V2, pre-trained using the Fully Convolutional Masked Autoencoder (FCMAE) framework, suitable for image classification and feature extraction tasks.
Downloads 485
Release Time : 1/5/2023

Model Overview

This is a lightweight convolutional neural network model primarily used for image feature extraction and classification tasks. It employs self-supervised learning for pre-training, without a pre-trained head, making it suitable for fine-tuning or use as a feature extractor.

Model Features

Lightweight design
The model has only 3.4 million parameters and requires 0.6 GMACs of computation, making it suitable for resource-constrained environments.
Self-supervised pre-training
Uses the Fully Convolutional Masked Autoencoder (FCMAE) framework for pre-training, eliminating the need for large amounts of labeled data.
Versatile feature extraction
Supports various usage modes, including image classification, feature map extraction, and image embedding.

Model Capabilities

Image feature extraction
Image classification
Generating image embeddings

Use Cases

Computer vision
Image classification
Classify images to identify the main objects within them.
Can be fine-tuned on the ImageNet-1k dataset to achieve classification capabilities
Feature extraction
Extract multi-level feature representations from images for downstream tasks.
Can output feature maps at different scales, suitable for tasks like object detection and segmentation
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase