# 無法使AMB82使用的SD卡內yolov7-tiny模型

**URL:** <https://forum.amebaiot.com/t/amb82-sd-yolov7-tiny/4550>\
**Category:** AI\
**Tags:** evb-amb82\
**Created:** [November 24, 2025, 10:05am UTC](https://forum.amebaiot.com/t/amb82-sd-yolov7-tiny/4550 "2025-11-24T10:05:52Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![ian1021222](https://avatars.discourse-cdn.com/v4/letter/i/4af34b/32.png) [@ian1021222](https://forum.amebaiot.com/u/ian1021222)\
**Post date:** [November 24, 2025, 10:05am UTC](https://forum.amebaiot.com/t/amb82-sd-yolov7-tiny/4550/1 "2025-11-24T10:05:52Z")

</div>

非常抱歉於這裡打擾，我現在使用4.08版本的開發環境，依照之前官方論壇討論中的幫助我在kaggle上使用yolov7-tiny訓練模型並要將其\*\*reparameterization(重參數化?)，\*\*但是每當放入我的best.pt時總會出現KeyError: ‘model.77.im.0.implicit’，我有試著在訓練模型時將IDect轉Dect使出來的模型結構已經由RepVGG 轉換參數化過了，但是換成.nb檔上傳板子後，板子仍然使用default的80個物品判斷模型，我把objectlist照教學改了，也把ObjDet.modelSelect(OBJECT\_DETECTION,CUSTOMIZED\_YOLOV7TINY , NA\_MODEL, NA\_MODEL);改上去了，仍然無法使用，想問一下有人有遇過這種問題嗎，我已參考過ruko2000在github上的方式無權照做了也無法完成，拜託幫幫我，感謝各位

---

<div class="post-metadata">

**Author:** ![KevinKL](https://sea2.discourse-cdn.com/flex016/user_avatar/forum.amebaiot.com/kevinkl/32/2632_2.png) [@KevinKL](https://forum.amebaiot.com/u/KevinKL)\
**Post date:** [November 25, 2025, 2:05am UTC](https://forum.amebaiot.com/t/amb82-sd-yolov7-tiny/4550/2 "2025-11-25T02:05:25Z")

</div>

您好

請問您是參考這份 [Kaggle](https://www.kaggle.com/code/kevinl00/yolov7-reparameterization-only) 來進行reparameterization 嗎? 有確認過nc參數的設定是否正確?

感謝

---

<div class="post-metadata">

**Author:** ![ian1021222](https://avatars.discourse-cdn.com/v4/letter/i/4af34b/32.png) [@ian1021222](https://forum.amebaiot.com/u/ian1021222)\
**Post date:** [November 25, 2025, 5:31am UTC](https://forum.amebaiot.com/t/amb82-sd-yolov7-tiny/4550/3 "2025-11-25T05:31:29Z")

</div>

是的，我的參數有調成1

---

<div class="post-metadata">

**Author:** ![ian1021222](https://avatars.discourse-cdn.com/v4/letter/i/4af34b/32.png) [@ian1021222](https://forum.amebaiot.com/u/ian1021222)\
**Post date:** [November 25, 2025, 6:34am UTC](https://forum.amebaiot.com/t/amb82-sd-yolov7-tiny/4550/4 "2025-11-25T06:34:23Z")

</div>

我試過有一位在論壇上用kaggle出模型判斷pothole並且成功跑過，我用了他的output完全照做，一樣無法確定AMB82是否用了SD卡內的模型，每次只要一跑，只要鏡頭對到任何人臉或物品，就會跳[ERROR] NNObjectDetection init failed: 0 172 0 121  
Item 1 pothole: 0 551 26 752Item 3 : 0 551 56 782 Item 6 � p� p� p

然後就跳出Bus Fault: SCB Configurable Fault Status Reg = 0x00000400，就跳出來了

---

<div class="post-metadata">

**Author:** ![ian1021222](https://avatars.discourse-cdn.com/v4/letter/i/4af34b/32.png) [@ian1021222](https://forum.amebaiot.com/u/ian1021222)\
**Post date:** [November 27, 2025, 9:52am UTC](https://forum.amebaiot.com/t/amb82-sd-yolov7-tiny/4550/5 "2025-11-27T09:52:58Z")

</div>

狀態更新:搞定了，yolov7-tiny請用最原始那位寫的程式以及權重([GitHub - WongKinYiu/yolov7: Implementation of paper - YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors](https://github.com/WongKinYiu/yolov7))就可以跑了，感謝各位，([YOLOv7 RefrigeratoryFood | Kaggle](https://www.kaggle.com/code/ianhungching/yolov7-refrigeratoryfood))這裡有kaggle的訓練，best.pt請到run這個資料夾裡找