# AI **URL:** https://forum.amebaiot.com/c/arduino/ai/124.md [Latest](https://forum.amebaiot.com/latest.md) · [Categories](https://forum.amebaiot.com/categories.md) · [Tags](https://forum.amebaiot.com/tags.md) --- ## [About the AI category](https://forum.amebaiot.com/t/about-the-ai-category/4527)
**Author:** [@M-ichae-l](https://forum.amebaiot.com/u/M-ichae-l)\ **Replies:** 0\ **Last updated:** [November 17, 2025, 4:00am UTC](https://forum.amebaiot.com/t/about-the-ai-category/4527 "2025-11-17T04:00:42Z")
For all topics related to Arduino SDK AI solution: Yolov3-tiny, Yolov4-tiny, Yolov7-tiny, NanoDet-Plus, SCRFD, MobileFaceNet, YAMNet, etc. 關於 Arduino SDK 人工智能方案: Yolov3-tiny, Yolov4-tiny, Yolov7-tiny, NanoDet-Plus, SCRF… --- ## [Send Audio to gemma4:e2b](https://forum.amebaiot.com/t/send-audio-to-gemma4-e2b/5075)
**Author:** [@rkuo2000](https://forum.amebaiot.com/u/rkuo2000)\ **Replies:** 8\ **Last updated:** [September 30, 2026, 9:05am UTC](https://forum.amebaiot.com/t/send-audio-to-gemma4-e2b/5075 "2026-09-30T09:05:52Z")
Application : AI plush toy LLM Server : EdgeAI-Gallery App with gemma4-e2b (on Android phone) EVB : AMB82-mini APK: EdgeAI-Gallery .apk , EdgeAI-Gallery-1.0.19.zip added text/image/audio API by codex. Code: Ollama\_s… --- ## [AMB82 線上模型轉換問題](https://forum.amebaiot.com/t/amb82/5088)
**Author:** [@nagual1414](https://forum.amebaiot.com/u/nagual1414)\ **Replies:** 2\ **Last updated:** [September 25, 2026, 11:58am UTC](https://forum.amebaiot.com/t/amb82/5088 "2026-09-25T11:58:41Z")
請問目前Online Conversion 是不是有問題呢? 昨天與今天嘗試上傳yolov7-tiny.pt 模型做轉換, 步驟有參考Youtube 影片教學. 但是當我按下SUBMIT 按鈕後, 過幾秒並沒有像之前一樣跳轉網頁, 而是清空剛剛所填的資料. --- ## [Send Image to gemma4:e2b](https://forum.amebaiot.com/t/send-image-to-gemma4-e2b/5071)
**Author:** [@rkuo2000](https://forum.amebaiot.com/u/rkuo2000)\ **Replies:** 3\ **Last updated:** [September 19, 2026, 3:29pm UTC](https://forum.amebaiot.com/t/send-image-to-gemma4-e2b/5071 "2026-09-19T15:29:31Z")
Application : Android Edge-Gallery App (Model = gemma4:e2b , AI Chat with Text/Image API) Example: Ollama\_sendImage.ino Question: the following code’s waitTime is fix at 10000 ms, how to modify it to keep waiting till … --- ## [Custom YOLOv7-tiny (2 classes) ghost boxes at top-left after Docker conversion — model verified correct pre-conversion (similar to #4667)](https://forum.amebaiot.com/t/custom-yolov7-tiny-2-classes-ghost-boxes-at-top-left-after-docker-conversion-model-verified-correct-pre-conversion-similar-to-4667/5029)
**Author:** [@navneet8875580017](https://forum.amebaiot.com/u/navneet8875580017)\ **Replies:** 5\ **Last updated:** [September 9, 2026, 10:25am UTC](https://forum.amebaiot.com/t/custom-yolov7-tiny-2-classes-ghost-boxes-at-top-left-after-docker-conversion-model-verified-correct-pre-conversion-similar-to-4667/5029 "2026-09-09T10:25:43Z")
Hi Realtek team, I’m deploying a custom-trained YOLOv7-tiny (2 classes: animal, person) on AMB82-Mini (RTL8735B), and getting the “ghost boxes clustered at top-left” issue described in thread #4667: I have systematica… --- ## [Agent MCP to interface with AMB82-mini](https://forum.amebaiot.com/t/agent-mcp-to-interface-with-amb82-mini/4831)
**Author:** [@rkuo2000](https://forum.amebaiot.com/u/rkuo2000)\ **Replies:** 3\ **Last updated:** [August 3, 2026, 6:30am UTC](https://forum.amebaiot.com/t/agent-mcp-to-interface-with-amb82-mini/4831 "2026-08-03T06:30:19Z")
Hi, For AI Agentic applications using AMB82-mini (such as intelligent lamp, robocar, AIoT device, considering to use Camera, Microphone, IMU, IR sensor) the agent could be using opencode or openclaw to interface with … --- ## [Converting EuroSAT ResNet18 to NN model](https://forum.amebaiot.com/t/converting-eurosat-resnet18-to-nn-model/4960)
**Author:** [@rkuo2000](https://forum.amebaiot.com/u/rkuo2000)\ **Replies:** 1\ **Last updated:** [July 23, 2026, 10:05am UTC](https://forum.amebaiot.com/t/converting-eurosat-resnet18-to-nn-model/4960 "2026-07-23T10:05:58Z")
Hi, I have trained EuroSAT ResNet18 model How do I convert the .pt file into NN model ? Your answer will be very appreciated ! --- ## [請問amb82-mini可以做語音辨識嗎?](https://forum.amebaiot.com/t/amb82-mini/3754)
**Author:** [@jordanhsu62](https://forum.amebaiot.com/u/jordanhsu62)\ **Replies:** 16\ **Last updated:** [July 4, 2026, 10:32pm UTC](https://forum.amebaiot.com/t/amb82-mini/3754 "2026-07-04T22:32:06Z")
請問amb82-mini可以做語音辨識嗎? 謝謝 --- ## [AMB82-Mini IoT AI Camera Arduino Dev. board RTSP streaming latency improvements?](https://forum.amebaiot.com/t/amb82-mini-iot-ai-camera-arduino-dev-board-rtsp-streaming-latency-improvements/1975)
**Author:** [@Benjamin\_Gatti](https://forum.amebaiot.com/u/Benjamin_Gatti)\ **Replies:** 5\ **Last updated:** [July 2, 2026, 7:20am UTC](https://forum.amebaiot.com/t/amb82-mini-iot-ai-camera-arduino-dev-board-rtsp-streaming-latency-improvements/1975 "2026-07-02T07:20:15Z")
Hi, I am using the public SDK, and I am wondering if H.264 can be configured for low latency by using settings such as zero-latency and sliced-based encoding on this chip/sdk? Thanks --- ## [請教 RTL8375B 的 NPU 多模型串接 (YOLOv7-tiny + CNN) 記憶體限制](https://forum.amebaiot.com/t/rtl8375b-npu-yolov7-tiny-cnn/4932)
**Author:** [@Casper\_Tu](https://forum.amebaiot.com/u/Casper_Tu)\ **Replies:** 1\ **Last updated:** [June 30, 2026, 6:28am UTC](https://forum.amebaiot.com/t/rtl8375b-npu-yolov7-tiny-cnn/4932 "2026-06-30T06:28:33Z")
各位工程師好，我看到官方支援 scrfd/mobilefacenet 的雙層人臉辨識串接。 如果我想要實作車牌辨識，打算在 NPU 上同時載入以下兩個模型做串接推論： yolov7-tiny (Object Detection)：用來偵測車牌外框。 CNN Gray/RGB (Image Classification)：用來辨識裁切下來的單一字母。 想請問： AMB82-MINI 的 RAM/NPU 記憶體，能夠同時… --- ## [在 Arduino IDE 實作 AudioClassification 時的 not align to 0x20 錯誤求助](https://forum.amebaiot.com/t/arduino-ide-audioclassification-not-align-to-0x20/4919)
**Author:** [@rick-kh](https://forum.amebaiot.com/u/rick-kh)\ **Replies:** 3\ **Last updated:** [June 24, 2026, 5:59am UTC](https://forum.amebaiot.com/t/arduino-ide-audioclassification-not-align-to-0x20/4919 "2026-06-24T05:59:50Z")
大家好，我是語音辨識開發的初初入門使用者，先感謝大家的協助。 我在 Edge Impulse 網站訓練了關鍵詞語音模型，下載 Quantized (int8) 或 Unoptimized (float32) 的 .tflite 檔並使用 Ameba\_acuity\_toolkit 順利產生 .nb 檔， 但是在 Arduino IDE 中使用 AudioClassification 範例來測試載入 .nb 檔時，Serial Moni… --- ## [How to trigger AI inference for X cycles](https://forum.amebaiot.com/t/how-to-trigger-ai-inference-for-x-cycles/4872)
**Author:** [@PingTKP](https://forum.amebaiot.com/u/PingTKP)\ **Replies:** 4\ **Last updated:** [June 8, 2026, 7:31am UTC](https://forum.amebaiot.com/t/how-to-trigger-ai-inference-for-x-cycles/4872 "2026-06-08T07:31:57Z")
Hi all, I would like to know whether it is possible to manually start image classification using the camera and stop it after it has run for X amount of times. Thanks in advance for the help! --- ## [ Request for Acuity Toolkit Access - GitHub: sacihanserdar-web](https://forum.amebaiot.com/t/request-for-acuity-toolkit-access-github-sacihanserdar-web/4885)
**Author:** [@serdar\_sacihan](https://forum.amebaiot.com/u/serdar_sacihan)\ **Replies:** 1\ **Last updated:** [June 2, 2026, 1:35am UTC](https://forum.amebaiot.com/t/request-for-acuity-toolkit-access-github-sacihanserdar-web/4885 "2026-06-02T01:35:23Z")
Hi, I am working on a license plate recognition project using AMB82-Mini. I have trained a YOLOv7-Tiny model and need Acuity Toolkit access for ONNX to NB conversion. GitHub: sacihanserdar-web Email: sacihanserdar@gmail.… --- ## [MobileNetV2 offline conversion](https://forum.amebaiot.com/t/mobilenetv2-offline-conversion/4849)
**Author:** [@PingTKP](https://forum.amebaiot.com/u/PingTKP)\ **Replies:** 2\ **Last updated:** [May 18, 2026, 8:16am UTC](https://forum.amebaiot.com/t/mobilenetv2-offline-conversion/4849 "2026-05-18T08:16:27Z")
Hi all, I am currently using Tensorflow v2.14.1 and Keras to train my model with MobileNetV2 as the base model. The model is as shown: \`\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_ Layer (type) … --- ## [Camera\_2\_Lcd\_JPEGDEC.ino error/warning](https://forum.amebaiot.com/t/camera-2-lcd-jpegdec-ino-error-warning/4811)
**Author:** [@rkuo2000](https://forum.amebaiot.com/u/rkuo2000)\ **Replies:** 6\ **Last updated:** [April 21, 2026, 1:11pm UTC](https://forum.amebaiot.com/t/camera-2-lcd-jpegdec-ino-error-warning/4811 "2026-04-21T13:11:40Z")
AMB82-mini , Arduion IDE 2.3.8, AmebaPro2\\4.1.1-build20260402 camera\_2\_lcd.ino compiled error Failed to open amebapro2\_fwfs\_nn\_models.json: 3 Failed to open amebapro2\_fwfs\_nn\_models.json: 3 Failed to open directory: … --- ## [AmebaPro2 AI convert Unauthorized Activity Detected](https://forum.amebaiot.com/t/amebapro2-ai-convert-unauthorized-activity-detected/4807)
**Author:** [@kuan](https://forum.amebaiot.com/u/kuan)\ **Replies:** 3\ **Last updated:** [April 8, 2026, 9:14am UTC](https://forum.amebaiot.com/t/amebapro2-ai-convert-unauthorized-activity-detected/4807 "2026-04-08T09:14:12Z")
I successfully converted the CNN model to network.nb, but when I clicked the download link, I encountered the following error. May I ask if there is something wrong with the website? (Previously, clicking it would downlo… --- ## [Inquiry regarding MobileNetV2 deployment workflow](https://forum.amebaiot.com/t/inquiry-regarding-mobilenetv2-deployment-workflow/4792)
**Author:** [@t114c75035](https://forum.amebaiot.com/u/t114c75035)\ **Replies:** 8\ **Last updated:** [April 2, 2026, 10:27am UTC](https://forum.amebaiot.com/t/inquiry-regarding-mobilenetv2-deployment-workflow/4792 "2026-04-02T10:27:59Z")
Hi Team, I am currently working on deploying a MobileNetV2 model to the Pegasus NPU platform. I have prepared the .pt weight file and a conversion script. I would like to verify if my workflow is correct, especially whe… --- ## [RGB\_CNN 模型的使用規則](https://forum.amebaiot.com/t/rgb-cnn/3713)
**Author:** [@josh](https://forum.amebaiot.com/u/josh)\ **Replies:** 8\ **Last updated:** [March 20, 2026, 1:49am UTC](https://forum.amebaiot.com/t/rgb-cnn/3713 "2026-03-20T01:49:22Z")
各位好， 我想請教關於 AMB82 是否支援自訓練的 CNN 分類模型的問題。如果可以使用，是否有特定的模型格式或類型需要遵循？ 我在官網上的文件中看到，AMB82 支援 CNN 模型的描述（官網說明），但同時也在論壇中發現提到 AMB82 不支援 CNN 模型的說法（論壇討論）。這讓我對 AMB82 的 CNN 支援能力感到困惑。 此外，我嘗試使用 MobileNetV3 訓練了一個分類模型，並透過官網推薦的工具進行模型轉換，但… --- ## [YOLO Survellience to notify user](https://forum.amebaiot.com/t/yolo-survellience-to-notify-user/4785)
**Author:** [@rkuo2000](https://forum.amebaiot.com/u/rkuo2000)\ **Replies:** 0\ **Last updated:** [March 16, 2026, 11:33am UTC](https://forum.amebaiot.com/t/yolo-survellience-to-notify-user/4785 "2026-03-16T11:33:31Z")
Hi, For my AI survellience project, I am using YOLOv7 to detect person, bicycle, car, motorbike, bus, truck and save image to SDcard by NTPclient timestamp (as filename) successfully. But I need notification to my W… --- ## [YOLOv7 Sushi Detection (renew)](https://forum.amebaiot.com/t/yolov7-sushi-detection-renew/4720)
**Author:** [@rkuo2000](https://forum.amebaiot.com/u/rkuo2000)\ **Replies:** 0\ **Last updated:** [February 10, 2026, 9:14pm UTC](https://forum.amebaiot.com/t/yolov7-sushi-detection-renew/4720 "2026-02-10T21:14:38Z")
For your reference, here’s the YOLOv7 trained by Dataset : Roboflow Sushi¶ --- ## [YOLOv7 Refrigatory Feed Detection](https://forum.amebaiot.com/t/yolov7-refrigatory-feed-detection/4719)
**Author:** [@rkuo2000](https://forum.amebaiot.com/u/rkuo2000)\ **Replies:** 0\ **Last updated:** [February 10, 2026, 9:11pm UTC](https://forum.amebaiot.com/t/yolov7-refrigatory-feed-detection/4719 "2026-02-10T21:11:02Z")
For your reference, here’s the YOLOv7 trained by Roboflow Dataset Refrigatory Food --- ## [YOLOv7 Face Mask Detection](https://forum.amebaiot.com/t/yolov7-face-mask-detection/4718)
**Author:** [@rkuo2000](https://forum.amebaiot.com/u/rkuo2000)\ **Replies:** 0\ **Last updated:** [February 10, 2026, 9:07pm UTC](https://forum.amebaiot.com/t/yolov7-face-mask-detection/4718 "2026-02-10T21:07:15Z")
For your reference, here’s the YOLOv7 trained by Kaggle Dataset Face Mask Detection --- ## [YOLOv7 Tanks Detection](https://forum.amebaiot.com/t/yolov7-tanks-detection/4717)
**Author:** [@rkuo2000](https://forum.amebaiot.com/u/rkuo2000)\ **Replies:** 0\ **Last updated:** [February 10, 2026, 9:00pm UTC](https://forum.amebaiot.com/t/yolov7-tanks-detection/4717 "2026-02-10T21:00:40Z")
For your reference, here’s the YOLOv7 trained by Roboflow Tanks --- ## [YOLOv7 Traffic Sign Detection](https://forum.amebaiot.com/t/yolov7-traffic-sign-detection/4708)
**Author:** [@rkuo2000](https://forum.amebaiot.com/u/rkuo2000)\ **Replies:** 0\ **Last updated:** [February 9, 2026, 1:26pm UTC](https://forum.amebaiot.com/t/yolov7-traffic-sign-detection/4708 "2026-02-09T13:26:49Z")
For your reference, here’s the YOLOv7 trained by Roboflow Traffic Sign v3i dataset https://www.kaggle.com/code/rkuo2000/yolov7-trafficsign --- ## [NPU Inference Discrepancy: Local conversion works, but Docker conversion results in "Ghost Boxes" at top-left](https://forum.amebaiot.com/t/npu-inference-discrepancy-local-conversion-works-but-docker-conversion-results-in-ghost-boxes-at-top-left/4667)
**Author:** [@t114c75035](https://forum.amebaiot.com/u/t114c75035)\ **Replies:** 2\ **Last updated:** [January 29, 2026, 7:47am UTC](https://forum.amebaiot.com/t/npu-inference-discrepancy-local-conversion-works-but-docker-conversion-results-in-ghost-boxes-at-top-left/4667 "2026-01-29T07:47:07Z")
I am attempting to package my NPU conversion toolchain (PyTorch to .nb) into a Docker container. Input Model: YOLOv7 .pt file (already reparameterized , already fix Torch 2.6 loading issue in in yolo/models/experiment… --- ## [詢問關於AMB82 AI 模型轉換](https://forum.amebaiot.com/t/amb82-ai/4663)
**Author:** [@Davis](https://forum.amebaiot.com/u/Davis)\ **Replies:** 3\ **Last updated:** [January 29, 2026, 2:45am UTC](https://forum.amebaiot.com/t/amb82-ai/4663 "2026-01-29T02:45:36Z")
最近嘗試在edge impulse製作模型並把模型使用在amb82-mini上，根據我所做出來的模型選擇的是CNN-GRAY並依據網站所提供的轉換要求(https://www.amebaiot.com/zh/amebapro2-ai-convert-model/)將.h5以及.tensor檔案上傳，結果有傳成功，但都沒收到回覆 :smiling\_face\_with\_tear: 。 這邊想問問有沒有人有相關經驗 … --- ## [Support convert Yamnet .h5 to .nb model](https://forum.amebaiot.com/t/support-convert-yamnet-h5-to-nb-model/4654)
**Author:** [@Hai\_Tran](https://forum.amebaiot.com/u/Hai_Tran)\ **Replies:** 5\ **Last updated:** [January 26, 2026, 6:20am UTC](https://forum.amebaiot.com/t/support-convert-yamnet-h5-to-nb-model/4654 "2026-01-26T06:20:20Z")
Hello Friends, I built a transfer learning Yamnet model below and saved as .h5 model. Then I used AI convert online to nb model, but I did not receive any result via email. I guess there is something wrong during conver… --- ## [Request: Offline AI Model Conversion Tool Access for Traffic Sign Detection POC](https://forum.amebaiot.com/t/request-offline-ai-model-conversion-tool-access-for-traffic-sign-detection-poc/4634)
**Author:** [@krish\_tyger](https://forum.amebaiot.com/u/krish_tyger)\ **Replies:** 3\ **Last updated:** [January 9, 2026, 9:35am UTC](https://forum.amebaiot.com/t/request-offline-ai-model-conversion-tool-access-for-traffic-sign-detection-poc/4634 "2026-01-09T09:35:54Z")
Hello, I’m working on a traffic sign detection POC using AMB82-MINI and need access to the offline AI model conversion toolkit (Acuity Toolkit). What I need: Access to ameba-ai-offline-toolkit GitHub repository Acuit… --- ## [詢問8735u結合生成式AI問題](https://forum.amebaiot.com/t/8735u-ai/4633)
**Author:** [@aa0900196900](https://forum.amebaiot.com/u/aa0900196900)\ **Replies:** 1\ **Last updated:** [January 2, 2026, 9:24am UTC](https://forum.amebaiot.com/t/8735u-ai/4633 "2026-01-02T09:24:03Z")
我想請問一下使用生成式AI的GeminiAI用一次就不能用了是有地方需要改嗎（？ --- ## [請問生成式AI使用gemini 的問題](https://forum.amebaiot.com/t/ai-gemini/4615)
**Author:** [@jordanhsu62](https://forum.amebaiot.com/u/jordanhsu62)\ **Replies:** 4\ **Last updated:** [December 31, 2025, 4:00am UTC](https://forum.amebaiot.com/t/ai-gemini/4615 "2025-12-31T04:00:10Z")
Amb82生成式AI使用gemini 今天才剛用就回復 You exceeded your current quota 不應該阿! 不是有免費的額度? gemini-2.0-flash-001不能用了嗎? 我試了較新的版本也不行. 可否給些建議 謝謝 [Next page](https://forum.amebaiot.com/c/arduino/ai/124.md?page=1)