Teaching Machines to Smell: New York Smells Dataset
What if computers could not only see, but also smell? Researchers introduce New York Smells, a large, real-world dataset pairing photos with olfactory sensor readings.
The set spans 7,000 smell-image pairs from 3,500 different objects across indoor and outdoor environments—about 70x more objects than prior smell datasets. That scale opens the door to training AI that understands the world through scent.
- Match a smell to the right photo (cross-modal retrieval)
- Recognize scenes, objects, and materials from smell alone
- Tell apart closely related scents—like different grass species
Early results: visual data helps machines learn richer scent representations, and these learned features beat widely used hand-crafted ones.
Smell is a core sense for animals. This work brings it closer to machines.
Dive into the paper: https://arxiv.org/abs/2511.20544v1
Paper: https://arxiv.org/abs/2511.20544v1
Register: https://www.AiFeta.com
#AI #MachineLearning #ComputerVision #Robotics #Olfaction #Multimodal #Datasets #Research
