-
|
Hello, When I am running my 1 channel models on my pc application code (QT application) using ncnn, results are fine. In android, I have a shader code which converts Bitmap from BGRA (little endia) to Y channel and which I am feeding to ncnn model like this void *data = (void *)env->GetDirectBufferAddress(image_buffer); and then I am returning results like this and after results are return my code is converting Y back to BGRA to draw on app.
you can see whole white background is looking like a net with holes. Note: I am using same shader and same model with tensor-flow lite on same application and results are fine. P.s. I tried 3 channel model also when I feed data using Please guide Thank you in advance |
Beta Was this translation helpful? Give feedback.
Replies: 3 comments
-
|
This fixed the issue |
Beta Was this translation helpful? Give feedback.
-
|
I made several attempts and ended up not proceeding, however I really wanted to test my trained model running on android with ncnn. I made an export on windows of a yolo11 model trained in colab. To test the converted model yolo --> ncnn I did a test with an image with two objects. C:\Users\guips\Downloads>yolo predict model=C:\Users\guips\Downloads\best_ncnn_model source="C:\Users\guips\Downloads\brasil\test129.jpg" save_txt=true image 1/1 C:\Users\guips\Downloads\brasil\test129.jpg: 640x640 2 Car, 103.0ms E/libEGL (14031): called unimplemented OpenGL ES API Build fingerprint: 'google/sdk_gphone64_x86_64/emu64xa:16/BP22.250221.010/13193326:user/release-keys' These are the first layers of the .param file and the last ones. 7767517 Last layer Split splitncnn_27 1 2 313 314 315 I thank you in advance for your attention and help. |
Beta Was this translation helpful? Give feedback.
-
|
I'm testing the use of ncnn on android. I converted a yolov11 model. I've tried every way to adjust the code, but I can't. The detections are not coming out correctly. The ncnn model on windows works as can be seen in the image. I'm sharing my native code in c++. If anyone can help me I'd really appreciate it. Image detected with yolo predict model=C:\Users\guips\Downloads\best_ncnn_model source="C:\Users\guips\Downloads\brasil\test822.jpg" save_txt=true Code C++ #include <jni.h> #define LOG_TAG "NCNN_DETECTOR" static ncnn::Net g_net; struct Detection { static float calculate_iou(const Detection& a, const Detection& b) { float inter_area = std::max(0.0f, x2 - x1) * std::max(0.0f, y2 - y1); return union_area > 0 ? inter_area / union_area : 0.0f; extern "C" JNIEXPORT jboolean JNICALL const char* param_path_str = env->GetStringUTFChars(param_path, nullptr); LOGI("Starting model loading..."); g_net.clear(); g_net.opt.num_threads = 4; int ret = g_net.load_param(param_path_str); ret = g_net.load_model(bin_path_str); FILE* fp = fopen(labels_path_str, "r"); char buffer[256]; if (g_labels.empty()) { LOGI("Model loaded successfully"); env->ReleaseStringUTFChars(param_path, param_path_str); extern "C" JNIEXPORT jstring JNICALL LOGI("Running updated detect method (no debug image)"); if (!g_model_loaded) { const char* image_path_str = env->GetStringUTFChars(image_path, nullptr); // Option 1: Use OpenCV (current) /* if (image.empty()) { LOGD("Input image: %dx%d, channels=%d", image.cols, image.rows, image.channels()); const int target_size = 640; float scale = std::min((float)target_size / img_w, (float)target_size / img_h); cv::Mat resized; cv::cvtColor(input_mat, input_mat, cv::COLOR_BGR2RGB); ncnn::Mat in = ncnn::Mat::from_pixels(input_mat.data, ncnn::Mat::PIXEL_RGB, target_size, target_size); // stbi_image_free(image_data); // Uncomment if using stb_image ncnn::Extractor ex = g_net.create_extractor(); if (out.empty()) { LOGD("Output dimensions: w=%d, h=%d, c=%d", out.w, out.h, out.c); int num_preds, num_classes; if (out.c == 1) { if (!is_decoupled) { if (num_classes != g_labels.size()) { std::vector detections; // Log raw outputs for (int i = 0; i < num_preds; i++) { } // NMS std::vector suppressed(detections.size(), false); for (size_t i = 0; i < detections.size(); i++) { } LOGD("Detections after NMS: %zu", final_detections.size()); // Format results return env->NewStringUTF(result_str.empty() ? "none" : result_str.c_str()); Here is a log of the output of the flutter application. [ ] D/NCNN_DETECTOR( 9180): Prediction 3730: x_center=640.00, y_center=640.00, width=640.00, height=640.00
|
Beta Was this translation helpful? Give feedback.
This fixed the issue
opt.use_fp16_packed = false;
opt.use_fp16_storage = false;
opt.use_fp16_arithmetic = false;