Whether it's extracting facial information in a vast crowd or identifying distant/small objects in a large image poses a major challenge to computer vision graphics. With years of technical accumulation, the research team from Carnegie Mellon University has finally found its flaws – the key to successfully identifying small objects is to find larger objects that match them.
This improved algorithm for extracting and encoding key information from pictures was promoted by Associate Professor Deva Ramanan and PhD student Peiyun Hu, and is a significant advance in identifying micro-human face miles.
In the face benchmark set, the previous method only identified between 29% and 64% of the correct human faces, and their improved algorithm reduced the privacy of the two errors, increasing the accuracy to 81. %.
Ramanan said: "It's like looking for a toothpick in someone's hand. It's very easy to see it when you suggest that the object may use a toothpick. The orientation of the finger, the movement and position of the hand are for us. The final finding of this toothpick provides a very important clue."
Similarly, in order to find faces with few pixels, the body or crowd photos in larger photos provide clues.
It has broad application prospects for the extraction of miniature faces, such as the number of statistical people. The demand for extension to micro-objects is becoming more and more prominent. When auto-driving cars are getting faster and faster, it is necessary to monitor and evaluate traffic conditions at all times. It is necessary to fully and correctly identify distant objects in order to make The correct response.
Ramanan said that it is not a new concept to help identify objects by assisting associated information. However, it is difficult to elaborate and express this intuition in practical systems. This is because the encoding of associated information usually involves "High-Dimensional Descriptors", which contains a lot of information but is very troublesome to use.
The method he and Hu developed was to use the Foveal Descriptors to simulate the human visual structure to encode the associated information. The fovea is the most sensitive area of ​​vision (color discrimination, resolution) in the retina. This method provides clear detail for small areas of the picture, while the surrounding area is more blurred.
By blurring the external image, the foveal description provides sufficient correlation information to help understand what is displayed in this area with a high degree of focus, and greatly reduces the computational burden. In this way, Hu and Ramanan's system is able to find and confirm the presence of faces in graphics blocks with fewer pixels.
Simply increasing the resolution of a picture may not be the best solution for finding tiny objects. With the high resolution, the problem of "Where"s Waldo is brought about. The target object contains a large number of pixels, which is very likely to be lost in the pixel. In this case, making full use of the associated information can help the system focus on the image block containing the face. In addition to the related information mentioned above, Ramanan and Hu indicate that it is very difficult to find a face in an image area having a small number of pixels by using a detector if the nose is detected several times in the same image area. As a result, they trained multiple independent detectors for different sized objects, greatly improving the ability to detect micro-objects.
The International Conference on Computer Vision and Pattern Recognition (CVPR 2017) will be held in Honolulu, the capital of Hawaii, USA from July 21st to 26th, and the research team will publish a detailed report. At present, the online version of the report has been released, and Lei Feng will continue to pay attention to the follow-up dynamics.
Silicone Purse
Silicone material with a smooth zipper that allows you to keep your essentials safe for backpacks,handbags,purses, Durability This bag will stay in shape for the time of its life. Made of 2 mm silicone, more resilient to punctures. Made with 100% BPA free silicone, is waterproof and fire resistant, also inhibits bacterial growth.Design Unique form created by double curved surfaces, roomy yet easy to hold. It stands up on it's own and fits well in your bag. Perfect for organizing makeup and toiletries.Ethical,Silicone is NOT a fossil fuel by product so it's more environment friendly.
Material:Silicone
Logo:Can custom design
Color:Any pantone color
MOQ:1000 pcs, Small quantity and Mix order are welcome
Safety:Nontoxic and eco-friendly anti-static, non-stick scraps and dust
Certificate:ROHS, REACH, LFGB, FDA
Packaging:Can Custom Polybag, Paper box, Metal box, Plastic box, Blister box(optional) 1PC/ OPP +insert card, Delicate packaging
OEM/ODM:OEM and ODM services provided Perfectly fit for every item
Sample time:5-7 DAYS
Payment:L/C,T/T, Western Union
Our main business is customization Silicone Phone Case,Silicone Rubber Sealing Ring Gaskets Grommets Accessories, Silicone Bracelets, Silicone Watchband, Silicone Keychains, Silicone Kitchenware. Silicone Spoon ,Silicone Stopper, Silicone Placemat Silicone Scrubber,Silicone Baby Products, Silicone Backpack ,Silicone Pet Supplies If you are interested, please consult
Our Factory Advantages:
1.Mold workshop and 2D and 3D engineer department
2.Solid siliccone compression machine and liquid silicone injection machine
3.Disney and Sedex 4P audit factory
4.ISO 9001,IATF16949,Raw material of FDA LFGB MSDS Certificates
Silicone Purse,Silicone Mini Purse,Waterproofing Coin Silicone Bag,Coin Purse Headset Bag
Shenzhen Yindingsheng Technology Co., Ltd , https://www.ydsoemsilicone.com