AIComputer vision’s shifting methods, from CNNs to transformers
A user’s retrospective describes transformers as dominant and reinforcement learning as rising, and points to 2024 as the biggest shift in computer-vision tasks.
TLDR
Looking back on computer vision in a September 20, 2026 post, a user says convolutional neural networks (CNNs) “came out of nowhere,” while recurrent neural networks and generative adversarial networks “had their moment.” They describe transformers as dominant and reinforcement learning as rising. For tasks, they identify 2024 as the year of the biggest change, with segmentation “losing the throne.”
Combined views
406
2 Sources, first seen 1h ago
Computer vision’s shifting methods, from CNNs to transformers
A user’s retrospective describes transformers as dominant and reinforcement learning as rising, and points to 2024 as the biggest shift in computer-vision tasks.