You May Also Enjoy
MMSA from beginner to proficient
less than 1 minute read
Published:
MMSA is a unified framework for multimodal emotion recognition developed by Tsinghua University. It supports three datasets, MOSI, MOSEI, and CH-SIMS, and 15 multimodal emotion analysis models, as well as tasks such as emotion recognition and emotion intensity regression. This column will use MMSA v1.0 as an example to implement framework analysis and customized modules.
Detectron2 Getting Started Tutorial
less than 1 minute read
Published:
Due to the high degree of encapsulation of Detectron2 and the obscure syntax, it often takes time to find the corresponding modules when carrying out the project. In addition, there is a lack of easy-to-understand introductory tutorials, and the official website tutorials lack flexibility. This series will start with topics such as Detectron2 installation, custom data sets, custom networks, and validation set loss printing to get started with Detectron2 step by step.
Openstl from zero to master
less than 1 minute read
Published:
Openstl is a third-party library developed by Westlake University for future frame prediction, which integrates multiple SOTA methods such as ConvLstm and simvip. This column will start from scratch and analyze the internal structure and customized modules of the entire Openstl project.
Classic Neural Network
less than 1 minute read
Published:
This blog series mainly introduces the traditional neural network structure design, including convolutional neural network, visual Transformer, Mamba, etc.