1. Hold a mid-term project inspection meeting, and communicate with team members and Dr. Wang Kun about the progress. Satisfied and accepted the overall project, and pointed out that the key parts need to be modified and improved and further developed to enhance the integrity of the system, functional highlights and interface quantity points. Specifically
①Pay attention to "highlight", that is, the layout highlights the theme function, such as the key function "text cross-modal search" query interface, highlighting the system-specific theme function, adding a frame with a sense of technology around the video playback interface, and adding functions on the edge (text search , video note positioning, time segment jump, etc.), a part is reserved on the right side (for example, to display the previous and next videos, etc., you can slide and choose to play), and the vertical version of the video needs to control the height to be smaller and the same as the horizontal one. Add a text search prompt box, connect to the text translation back-end interface, and change the video note from the corresponding time stamp to the corresponding time segment.
② Add system auxiliary functions: files, cloud disk, etc., pay attention to improving the number of function points and increasing the degree of perfection. Expand around the theme function.
③Currently existing problems: the return value of the unfinished cross-modal AI model interface is not only the Top5/10 List timestamp fragment, but also several values such as confidence score-presentation effect, which can be obtained through echarts or vue front-end technology Display the table and click to select the function of displaying video clips.
④The focus of the system is on the data set (minimum 3-5min/longest 30min) [The priority of repairing the problem of slow video upload is low, add sample video, resolution 480P/360P, size control within 100 megabytes] optional display compression, There are more features in the backend, clearer content, binding id in the backend, etc.
The meeting minutes and mid-term inspection are as follows:
2. Back-end database design and maintenance. The backend is currently compressed into three tables. The user table completes user personal information storage, the video table completes video information storage, and the video_note table completes video note fragment storage. The later text retrieval cross-modal table organization structure is similar to it, and the cache technology is constructed to optimize data storage.
3. The current project development progress. Based on the first version of the project, the layout, technological color matching, and function list have been optimized, and all functions (except text retrieval) have been realized.
