Read the article
Say it early. If the interviewer asks about the project and you describe it as finished, the follow-up questions will find the gap and it will look as though you were hiding it. Put it in your first answer instead: 'The detection works on our test images, but on live camera input it is too slow to be useful.' Now the conversation is about something real.
Explain the cause as you understand it now. Students often give the first explanation they had at the time. By the end of a project you usually know more. 'We thought it was the camera. It turned out the model was too large for the board, and we found that out too late to train a smaller one.' That sentence shows you can diagnose a problem, which is one of the things the interviewer is trying to find out.
Separate what was in your control from what was not. Components that arrived late, a lab that was closed, or a dataset that was withdrawn are real reasons and you can mention them once. Then spend your time on the parts you could have done differently. Interviewers listen for whether a candidate takes responsibility for their share without taking on blame that is not theirs.
Describe what you tried, in order. 'First we reduced the image size. That helped a little. Then we tried a smaller model, which was faster but missed too much. We ran out of time before trying the third idea.' This is the most valuable part of the answer, because it shows how you work when something is not going well, and that is a large part of any job.
Say what you would do differently, and make it specific. 'Test on the real device in the first month, not the last' is useful. 'Plan better' is not.
If the problem was a teammate, be careful. It is fine to say 'one member left the team in the sixth semester, so three of us split the work.' It is not fine to describe that person's faults. Interviewers notice how you speak about colleagues who are not in the room.
Finally, be ready to show what does work. Even a project that failed overall usually has a part that runs. Know it well, and offer to walk through it. The interviewer is less interested in whether the project succeeded than in whether you understand why it did or did not.