SK-TDA4VM: kit can't boot-up

All boards are tested at manufacturing, thus it seems highly unlikely you would receive two faulting boards.  What version of the board is failing and working?  There should be a sticker that says PROC112E?

What power supply are you using (same supply for both working/failing board)?

In the user guide, there is Section 5 Usage Notes and Advisories.  Do any of these apply to your setup?

"Object detection sample code which downloaded from you on that board, the results is all wrong".   Can you please provide additional information on this?

Both working and failing board’s sticker are PROC112A1.

Both of them use the same power supply and SD card to do all the testing.

I'm sorry I said it wrong, it’s face detection sample code not object detection.

The result of the sample code should draw bounding boxes on detected faces, but the drawn boxes looks draw randomly on image.

There's supposed to be one box(on my face) but he shows up in a bunch of boxes on everywhere.

Can you provide more info on "Every time I tried to reboot it by unplugging it, it wouldn't start unless leave it for a while."?

What type of environment is the board operating (indoor/lab)?  Is the temperature of the board getting elevated (heated) causing this issue?

It starts normally at first, but then it would suddenly get stuck, so I had to reboot it.

Because it was stuck, I had to pull the plug out, and when it was plugged back in, it was completely unresponsive.

The light on the board was on, but there was no startup screen on the screen.

At first I thought it was broken, but after a while I plugged in and there was boot screen.

Then I ran the face detection program directly, and it turned out just like I said last mail.

Then I realized that even when it wasn't stuck, couldn't turn it on if unplugged it, but it worked after leave it for a while.

There was a problem with the temperature, I was indoors, but it didn't take long to get that problem.

He did get hot, but not as hot as the other boards normally use.

Yes - sounds like the board is behaving strangely.  My only guess is that it is thermal related.  Maybe something is running hotter than normal, making the system more sensitive to heat?  Can you set a fan next to the board to get air moving across it constantly.  Does that help make more stable?

I'll test the fan part, but as I mentioned in my last mail, It has affected the accuracy by running the program when it's cold.

Maybe there are other problems?

Is there any way to find the problem as soon as possible, because I can't get the error log this time.

We've not experienced this issue from other customers (or internally) - so not sure what is causing the problem (thermal, power supply, faulty board, software changes, etc). 

You can't get error log - so board is working now?

The problem with the board when it works is that it can't detect the right thing and stuck soon.

When the board went wrong, it was completely stuck(no display), so I couldn't give you the error log like I did with the last board.

I'm sorry for your experience with the board's inconsistencies.  It does appear there is an issue with the board.  You can contact TI support and use this E2E thread as confirmation that you EVM does have a hardware problem and needs to be replaced

https://www.ti.com/productreturns/docs/createReturn.tsp.