Doc9 Posted December 11, 2020 Posted December 11, 2020 Greetings! Thank you in advance for reading my post. If you are willing/able to provide any assistance it would be greatly appreciated. Request: I am seeking a point of contact with expertise in video image processing to provide insight and guidance on appropriate data formatting for follow on neural network analytics. Background: I am a physician who is developing a research project that uses video data to derive diagnostic indicators. I have a deep understanding of the medical problem sets and a sense of what is possible using various emerging AI capabilities. However, I have significant technical knowledge gaps regarding the bridge between data collection and data analytics. I am trying to avoid the scenario where extensive funding and resources have been applied to collect the data and it is not optimized or feasible for analysis. Other: There are potential avenues for compensation and/or collaboration related to assistance with this project. Thank you!
fiveworlds Posted December 11, 2020 Posted December 11, 2020 Hello, the right data format would entirely depend on the data that is being collected.
Sensei Posted December 13, 2020 Posted December 13, 2020 Generally, graphics can be stored in still images, which can be losslessly compressed (e.g. PNG), or lossy compressed (JPEG). There are not many video lossless compression file formats. MPEG, MP4, AVI are typically lossy compressed. https://en.wikipedia.org/wiki/Lossless_compression https://en.wikipedia.org/wiki/Lossy_compression Digital cameras store JPEG still photos, or MP4 videos. Both lossy compression. On 12/11/2020 at 4:53 PM, Doc9 said: I am trying to avoid the scenario where extensive funding and resources have been applied to collect the data and it is not optimized or feasible for analysis. ..you should keep original raw photos and videos... just in case you will have to restart learning from scratch.. Single 10 Mpix JPEG photo takes ~2.5 MB, so 1 TB SSD (for $100) will be enough for 400,000 still photos.
Doc9 Posted December 16, 2020 Author Posted December 16, 2020 Thanks you Sensei and Fiveworlds. I appreciate your replies. I will definitely take a look at the links you posted, Sensei. No doubt a good place for me to start getting smart on this subject.
fiveworlds Posted December 17, 2020 Posted December 17, 2020 (edited) We have limited idea of what the content of the videos will be. We had a previous example where we can easily count individual birds but would have difficulty counting a flock of birds with touching wings. The resolution of the images is very important if you need to distinguish specific things. To give a medical example you could have a slide under the microscope of patient blood. But if the cells are too clumped together it can be difficult for the computer to match the shapes of the cells against known illnesses. Edited December 17, 2020 by fiveworlds
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now