TikTok's 'Depth Anything' model, a new state-of-the-art monocular depth estimation tool, outperforms MiDaS with its zero-shot capabilities across various datasets and real-world photos. The model, now integrated into the 'transformers' library and available on Hugging Face, is accessible for easy use and implementation. Despite its small size of only 25 million parameters, the lightweight version runs efficiently in the browser via Transformers.js, eliminating the need for a server. 'Depth Anything' was trained on 62 million images using a 'data engine' that adds more labeled data recursively. It delivers LiDAR quality depth estimation and can be used with just three lines of code, as demonstrated in a provided demo notebook. Notably, Lihe Yang, the first author of the model, developed it during an internship with the company.
Elon Wuz Right: Vision is All You Need > TikTok team trains Depth Anything > LiDAR quality depth estimation from single photo frame > Using teacher model-student model system > 1st author Lihe Yang did this while on his internship (!) with the company Out of training set… https://t.co/IRqjZ335h6
Depth Anything is now available @huggingface!! A DPT model trained on 62 million images for monocular depth estimation using a clever "data engine" leveraging lots of unlabeled data It can now be used in 3 lines of code! 🔥 Demo notebook: https://t.co/gG52yTUSSe https://t.co/h1C59SCXOI
Depth Anything is now available @huggingface, a model trained on 62 million images for monocular depth estimation using a clever "data engine", recursively adding more labeled data to the mix It can now be used in 3 lines of code! 🔥 Demo notebook: https://t.co/gG52yTUSSe https://t.co/F95t6YrQ9k
Depth Anything is now available in 🤗 Transformers.js! At just 25M parameters, the small version of the model runs great locally. Here's a demo I created which performs monocular depth estimation directly in your browser (no server needed)! 🤯 🔗 Demo: https://t.co/bMKDgkro3y https://t.co/2pZoqqVQVm
Explaining a new state-of-the-art monocular depth estimation model: Depth Anything ✨ 🧶 It has just been integrated in transformers for super-easy use. We compared it against DPTs and benchmarked it as well! You can the usage, benchmark, demos and more below 👇 https://t.co/k1EgRKlo7P
Zero-shot depth estimation that outperforms MiDaS 🤯 Meet Depth Anything by TikTok, a simple yet powerful foundation model for robust monocular depth estimation. It shows exceptional zero-shot capabilities across six public datasets and real-world photos, demonstrating… https://t.co/CxvCWumJUj