leftthomas / clipprompt Goto Github PK

Hello, I apologize for any previous lack of courtesy in my response. Regarding your question, here's a polite translation:

"Hello, I followed the instructions to set up the environment properly, but I encountered the following error when running the program: 'File "D:\Anaconda3\envs\lixue\lib\ctypes_init_.py", line 373, in init self._handle = _dlopen(self._name, mode) OSError: [WinError 127] The specified program could not be found.' Have you ever encountered this issue?"

Hello! I just wanted to start by saying that you've done an amazing job! I did notice, however, that while the mAP@200 achieved in your implementation (71.92) is very close to the value reported in the paper (72.38), there seems to be a significant difference in the P@200 metric between your code (68.05) and the paper's reported value (0.725). I was wondering if you had any insights regarding this discrepancy?