logo
Loading...

如何解決InvalidArgumentError: Invalid JPEG data or crop window錯誤 - Cupoy

合併之後, 更改train.record和test.record, 進行training. 會出現以...

如何解決InvalidArgumentError: Invalid JPEG data or crop window錯誤

2020/03/16 下午 07:55
電腦視覺深度學習討論版
Liaw Jiun
觀看數:14
回答數:1
收藏數:0

合併之後, 更改train.record和test.record, 進行training. 會出現以下錯誤碼:error message: InvalidArgumentError:  Invalid JPEG data or crop window查閱google後(下述連結), https://blog.csdn.net/qq_36190978/article/details/95937723?depth_1-utm_source=distribute.pc_relevant.none-task&utm_source=distribute.pc_relevant.none-task發現可能有些JPEG格式是progressive, 而不是baseline. 修改完,這些JPEG後, 檢查全部都是baseline後, 重train, 仍然出現相同error message: InvalidArgumentError:  Invalid JPEG data or crop window請問該如何解決????


2020/03/17:  發現JPEG改正後, 沒有重新產生train.record和test.record所導致. 但是run下去後, 發現下面的error code.  請問該如何解決??????


=== error code====

INFO:tensorflow:Loading and preparing annotation results... I0317 10:01:26.821242 140199887435520 coco_tools.py:115] Loading and preparing annotation results... INFO:tensorflow:DONE (t=0.01s) I0317 10:01:26.830586 140199887435520 coco_tools.py:137] DONE (t=0.01s) creating index... index created! 2020-03-17 10:01:26.839506: W tensorflow/core/framework/op_kernel.cc:1639] Invalid argument: TypeError: object of type <class 'numpy.float64'> cannot be safely interpreted as an integer. Traceback (most recent call last):    File "/usr/local/lib/python3.6/dist-packages/numpy/core/function_base.py", line 117, in linspace     num = operator.index(num)  TypeError: 'numpy.float64' object cannot be interpreted as an integer   During handling of the above exception, another exception occurred:   Traceback (most recent call last):    File "/tensorflow-1.15.0/python3.6/tensorflow_core/python/ops/script_ops.py", line 235, in __call__     ret = func(*args)


.....

....

回答列表

  • 2020/03/22 下午 00:00
    Jeffrey
    贊同數:0
    不贊同數:0
    留言數:5

    有幾個可能:

    在做DATA處理的時候, JPEG 有ERROR 也許是 bit loss.

    建議是從頭執行:

    1. 原始的資料集先行利用 open CV 確認所有 JPEG 都是 baseline

    2. DATA augument 後, 確認 IMAGE data size 符合

    3. 資料集分成 train/test

    4. 進行訓練