Default Avatar

Maria Chen

Maria Chen

About

Username
Maria Chen
Joined
Visits
1,112
Last Active
Roles
Member, ModeratorsKL520, ModeratorsKL720, ModeratorsAIModel

Comments

  • Hi Richard, 這個訊息很有可能是來自於硬體的限制。 如果您可以提供.nef檔，.onnx檔，還有轉檔的script，我們可以幫忙確認。
    in Yolov7轉檔問題 Comment by Maria Chen September 2022
  • Hi Anoop, It seems like this model hasn't gone through ONNX to ONNX. Please refer to Toolchain Manual - Document Center (kneron.com), sections 3.1.2 Pytorch to ONNX and 3.1.5 ONNX Optimization. After using the Python API to go through these 2 steps,…
    in Error in generating the nef file Comment by Maria Chen September 2022
  • Hi, 推論的圖片格式與模型轉換的圖片格式沒有關係，所以img_channel設定為RGB就可以了喔。 推論時，請將image format設定成KP_IMAGE_FORMAT_YUYV。 https://www.kneron.com/forum/uploads/306/WRPQ7FYUF9PU.png
    in kl520模型轉換問題 Comment by Maria Chen September 2022
  • Hi Anoop, Could you run "eval_result = km.evaluate()"? If so, what does the result say? There are some possible reasons, such as the input shape is incorrect, or there is an unsupported operator, but we would need more information to find …
    in Error in generating the nef file Comment by Maria Chen September 2022
  • Hi Dennis, Is the SDK version you're using KL520 SDK v1.7.0? Also, are you using OpenCV on your PC? If so, OpenCV's default color format is BGR, and the color format KL520 SDK uses is RGB565, so you might want to convert your image to BGR before dis…
    in Wrong color using USB to PC transmitted by DevKit Comment by Maria Chen September 2022
  • Hi Su, KL720DemoGenericInferenceYoloX_BypassHwPreProc.py裡面的postprocessing是給YoloX用的，所以可能會有問題。 要用任何模型，可以去耐能的開發中心下載Kneron PLUS (開發中心 | Kneron - 人工智慧無所不在)，有一些C與Python範例可以跑，像是KL720DemoGenericIm…
    in Run your NEF model and your image on KL720 USB accelerator 問題 Comment by Maria Chen September 2022
  • 您好， 太好了! 很高興訓練模型指令可以正常運作。 這個FileNotFoundError，是因為dataset的實際路徑是data/imagenet/val/O/O_xxx.jpg (我是在train的資料夾找到O_1004.jpg的)，圖片外面有O跟R的資料夾，所以這個路徑是錯的。 建議您可以把O與R的圖片，全都放在Val資料…
    in 已訓練模型在轉檔時遇到的種種問題 Comment by Maria Chen September 2022
  • 您好， 我們的產品沒有特別說支援哪些模型，不過我們有NPU支援的operator list，若模型中使用的都是NPU有支援的op，那就可以在KL520上面執行。 https://www.kneron.com/forum/uploads/319/94E0OU134SH3.png Operator list: Toolchain Manual - Document Center (kneron…
    in KL520支援YOLO版本 Comment by Maria Chen September 2022
  • 您好! 我現在下載您的dataset確認了，可以正常訓練 (但是Validate時會有FileNotFoundError)。 所以不是dataset的問題，不好意思! 在regnetx-400mf_8xb128_in1k.py裡，我只有改samples_per_gpu與workers_per_gpu，也有在kneron-mmclassification裡新增data與imagenet資…
    in 已訓練模型在轉檔時遇到的種種問題 Comment by Maria Chen September 2022
  • 您好， kp.GenericInageInferenceDescriptor是Kneron PLUS v.2.0用的，所以您可以用Kneron PLUS v1.3.0的kp.GenericRawImageHeader喔。 這裡有兩個版本的API對照表: V1.x and V2 APIs Mapping Table - Document Center (kneron.com) https://www.kneron.com/forum/uplo…
    in generic image inference input descriptor的kp.GenericImageInferenceDescriptory在Kneron Plus v1.3.0已不存在 Comment by Maria Chen September 2022
  • 您好， for filename, gt_label in samples: ValueError: not enough values to unpack (expected 2, got 1) 有可能是因為samples中，沒有分別拿到filename與gt_label (如下圖) https://www.kneron.com/forum/uploads/002/T1JIUF8T92C2.png 有兩種dataset格式符合架構…
    in 已訓練模型在轉檔時遇到的種種問題 Comment by Maria Chen September 2022
  • 您好， -Can't open file 'tools/deployment/pytorch2onnx.py' No such file or directory: 您可以在跑程式前，先用Colab的files view找到檔案，再用copy path複製路徑。Colab的路徑，與使用者mount的Google Drive資料夾相關，這個部分可以在網路上找到相關說明。 -Key…
    in 已訓練模型在轉檔時遇到的種種問題 Comment by Maria Chen September 2022
  • Hi, 這個問題如果在docker那邊問，可能會有比較好的答案，但是您可以試著跑跑看docker image prune，這會刪除沒有在用的container。或是docker rmi $(docker images -q --filter "dangling=true")，這會移除dangling images。 或者您可以確認當初安裝raspber…
    in raspberry pi4 Comment by Maria Chen September 2022
  • 您好, 您提到的project (host_lib)裡面也有postprocessing。dme_inference裡面會呼叫dme_get_result，在那裡面會拿raw output (feature map)去做postprocessing: https://www.kneron.com/forum/uploads/881/E2Q5VG38ZZUB.png Host_lib裡的postprocessing範例在host_lib…
    in 關於kl520輸出的post processing Comment by Maria Chen September 2022
  • Hi Johnson, 謝謝您提供的資訊! 我們看了您提供的程式。 在KL520上，需要先跑load_model_from_file (下面被註解掉的地方)，才能跑get_model_info，因為usb-boot upload fw會洗掉DDR上的所有資料(包含模型)。 https://www.kneron.com/forum/uploads/331/7ZP6BG0LFYKB.pn…
    in 無法上傳 MODEL 至 kl520 Comment by Maria Chen September 2022
  • Hi Johnson, 可以跟您請教一下您的OS版本嗎 (Windows, Linux, etc.)? 還有，您的KL520的USB加速棒是如何與您的電腦連接的呢? 您的電腦現在有連接著很多其他的東西嗎? 因為load model時會再次與KL520加速棒連接，這時如果連接速度太慢，有可能會造成這個error。 P.S. …
    in 無法上傳 MODEL 至 kl520 Comment by Maria Chen September 2022
  • Hi, 在執行gereric inference時不需要加sudo，因為在python環境中用sudo的話會跑到另外一個環境，可能就不是您安裝的環境了。 若是沒有用sudo會遇到USB權限問題，請參考文件說明設定使用權限: Install Dependency - Document Center (kneron.com)
    in raspberry pi4 Comment by Maria Chen August 2022
  • Hi Su, Tutorial在docker裡面，需要先安裝docker，這個文檔有安裝方式與連結: Toolchain Manual - Document Center (kneron.com) Docker的用法: http://doc.kneron.com/docs/#toolchain/manual/#2-toolchain-docker-overview Tutorial資料夾在docker裡的位置是: /works…
    in yolov5 Object Detection Comment by Maria Chen August 2022
  • Hi Richard, Error 111 是 KP_FW_INVALID_PRE_PROC_MODEL_INPUT_SIZE_111，當preprocessing的input size不符合規範的時候會有這個error。 因為像您所說的，2112比1023大，所以會有error。這部分是KL720目前inproc (hardware preprocess) 的限制，目前超過的話可以透過b…
    in KL720使用kneron_plus v2.0.1出現inference failed, error=111 Comment by Maria Chen August 2022
  • Hi Gopinath, No problem! I don't think there is. The examples for AI SoC Development Kit only uses the Tiny Yolo V3 model. What you could do is use Kneron PLUS' Generic Inference example in Python to inference with your own model: Run Examples - Doc…
    in Kneron PLUS - Python (Establishing device connection) Comment by Maria Chen August 2022
  • Hi Jack, 謝謝您提供的結果! 我們的同事有再調整一下您的 code ，趨近您理想的結果。 https://www.kneron.com/forum/uploads/477/RJLX9NHDGV2K.zip 數據不同的原因主要是您的inference header要送的buffer size沒有填正確 (原本送的是沒有alined過的data size) 名字裡…
    in Inference with not image data Comment by Maria Chen August 2022
  • Hi Jack, 這是我們同事跑出來的數據，請問一下這個有符合您預期的結果嗎? https://www.kneron.com/forum/uploads/690/N2BCPSF2V8FU.txt
    in Inference with not image data Comment by Maria Chen August 2022
  • Hi Jack, 第一個問題(module 'kp' has no attribute 'v1')，是因為版本不同(v1.3.0/v2.0.0)，所以需要再改程式裡的一些地方才能成功地跑。 第二個問題，轉成float的動作是只要用bypass_preprocess，都會需要去做。 下面的圖，Data Inference的地方有幾個步驟: https://…
    in Inference with not image data Comment by Maria Chen August 2022
  • Hi Jack, 要inference之前，我們會把float (32 bits)去做浮點到定點(8-bit fixed point)的轉換來量化: https://www.kneron.com/forum/uploads/531/VG6DDPP878BD.png 然後這裡(上面回覆的連結)有說明要如何排列資料，您的layout會需要用KP_MODEL_TENSOR_DATA_LAYOUT_1W1…
    in Inference with not image data Comment by Maria Chen August 2022
  • Hi Jack, There's also an example talking about inferencing none image data without built-in hardware image pre-processing here: Data Inference - Document Center (kneron.com) I hope this helps.
    in Inference with not image data Comment by Maria Chen August 2022
  • 您好: 不好意思，我不太清楚義晶的板子，可能需要請您詢問義晶的人員或是販賣這個AI BOX給您的人幫您處理了。
    in KL720功能實現問題 Comment by Maria Chen July 2022
  • Hi Gopinath, Are you talking about the KL520 series AI SoC Developement Kit? If so, there is an Assembly Instruction inside. The board should be set up like the photo below. If this wasn't what you were asking about, please let us know what KL520 S…
    in Kneron PLUS - Python (Establishing device connection) Comment by Maria Chen July 2022
  • Hi Jack, Sorry for the confusion, I'll try my best to explain this. Since you're bypassing the preprocess, you could convert your frames into an array first. Then, you would need to know your model's normalization, and write a function that converts…
    in Inference with not image data Comment by Maria Chen July 2022
  • Hi Gopinath, I'm glad I could help! And sure, you could download Kneron PLUS from here Developers | Kneron – Full Stack Edge AI to do inferencing with your camera on your laptop, or with a USB camera connected to your computer. If you use C, you co…
    in Kneron PLUS - Python (Establishing device connection) Comment by Maria Chen July 2022
  • 您好， 謝謝您的回饋，很高興程式可以執行成功!
    in app_face_detection_recognize發生韌體問題(KL520) Comment by Maria Chen July 2022

© Kneron, Inc.