I created a large‑scale paired dataset of over 6 000 photos captured simultaneously by a DSLR and three low‑end smartphones under diverse conditions, and trained a deep convolutional neural network to map phone images to DSLR‑like quality.
large-scale dataset of over 6K photos was collected synchronously by a DSLR camera and 3 low‑end cameras of smartphones in a wide variety of conditions.
A Deep Convolutional Networks AI based software which takes bad quality photos and turn them into really good quality photo by enhancing not just megapixels but all the attributes which differentiate a DSLR photo to a phone camera photo.
The proposed solution is a pair of AI‑powered AR glasses that locate personal items (e.g., keys) using computer‑vision, ego‑motion, SLAM, and neural‑network based perception while the wearer moves around the house.
- Samyak Datta
- Sameer Dharur
- Vincent Cartillier
- Ruta Desai
- Mukul Khanna
- Dhruv Batra
- Devi Parikh
I wrote this note on 2022-09-06 to share a quick checklist of iPhone 14‑era computer‑vision kit capabilities that are relevant for building everyday consumer vision apps on iOS devices.
- ✅ Contour detection as shown in the picture ✅ Detect and classify human activity in real time ✅ Locate and demarcate rectangles, faces, barcodes, and text in images ✅ Analyze and label images ✅ Analyze image similarity ✅ Applying matte effects to people in images and video ✅ Highlighting and cropping interesting areas of images using saliency ✅ Object tracking ✅ Select a selfie based on capture quality ✅ Animal detection 🤔 which ones and why???
The note ends with hashtags that capture the broader context: #computervision #iphone #datapreparation #edgecomputing #objectdetection #poseestimation #iosdeveloper #feedbackloop.