Output improvements for detector results: When printing detector results, o...

iovodov

2018-05-03 028696bf15efeca3acb3db8c42a96f7b9e0f55ff

Output improvements for detector results:
When printing detector results, output was done in random order, obfuscating results for interpreting. Now:
1. Text output includes coordinates of rects in (left,right,top,bottom in pixels) along with label and score
2. Text output is sorted by rect lefts to simplify finding appropriate rects on image
3. If several class probs are > thresh for some detection, the most probable is written first and coordinates for others are not repeated
4. Rects are imprinted in image in order by their best class prob, so most probable rects are always on top and not overlayed by less probable ones
5. Most probable label for rect is always written first
Also:
6. Message about low GPU memory include required amount

 src/box.h

@@ -30,6 +30,9 @@
    float *mask;
    float objectness;
    int sort_class;
    // The most probable class id: the best class index in this->prob.
    // Is filled temporary when processing results, otherwise not initialized
    int best_class;
} detection;

box float_to_box(float *f);

 src/convolutional_layer.c

@@ -491,7 +491,7 @@
    size_t total_byte;
    check_error(cudaMemGetInfo(&free_byte, &total_byte));
    if (l->workspace_size > free_byte || l->workspace_size >= total_byte / 2) {
        printf(" used slow CUDNN algo without Workspace! \n");
        printf(" used slow CUDNN algo without Workspace! Need memory: %d, available: %d\n", l->workspace_size, (free_byte < total_byte/2) ? free_byte : total_byte/2);
        cudnn_convolutional_setup(l, cudnn_smallest);
        l->workspace_size = get_workspace_size(*l);
    }

 src/image.c

@@ -229,27 +229,70 @@
    return alphabets;
}

void draw_detections_v3(image im, detection *dets, int num, float thresh, char **names, image **alphabet, int classes)
{
    int i, j;

    for (i = 0; i < num; ++i) {
        char labelstr[4096] = { 0 };
        int class_id = -1;
        for (j = 0; j < classes; ++j) {

// Creates array of detections with prob > thresh and fills best_class for them
detection** get_actual_detections(detection *dets, int dets_num, int classes, float thresh, int* selected_detections_num)
{
    int selected_num = 0;
    detection** result_arr = calloc(dets_num, sizeof(detection*));
    for (int i = 0; i < dets_num; ++i) {
        dets[i].best_class = -1;
        for (int j = 0; j < classes; ++j) {
            if (dets[i].prob[j] > thresh) {
                if (class_id < 0) {
                    strcat(labelstr, names[j]);
                    class_id = j;
                if (dets[i].best_class < 0 || dets[i].prob[dets[i].best_class] < dets[i].prob[j]) {
                    dets[i].best_class = j;
                }
                else {
                    strcat(labelstr, ", ");
                    strcat(labelstr, names[j]);
                }
                printf("%s: %.0f%%\n", names[j], dets[i].prob[j] * 100);
            }
        }
        if (class_id >= 0) {
        if (dets[i].best_class >= 0) {
            result_arr[selected_num] = &(dets[i]);
            ++selected_num;
        }
    }
    if (selected_detections_num)
        *selected_detections_num = selected_num;
    return result_arr;
}

// compare to sort detection** by bbox.x
int compare_by_lefts(const void *a, const void *b) {
    const float delta = ((*(detection**)a)->bbox.x - (*(detection**)a)->bbox.w / 2) - ((*(detection**)b)->bbox.x - (*(detection**)b)->bbox.w / 2);
    return delta < 0 ? -1 : delta > 0 ? 1 : 0;
}

// compare to sort detection** by best_class probability 
int compare_by_probs(const void *a_ptr, const void *b_ptr) {
    const detection* a = *(detection**)a_ptr;
    const detection* b = *(detection**)b_ptr;
    float delta = a->prob[a->best_class] - b->prob[b->best_class];
    return delta < 0 ? -1 : delta > 0 ? 1 : 0;
}

void draw_detections_v3(image im, detection *dets, int num, float thresh, char **names, image **alphabet, int classes)
{
    int selected_detections_num;
    detection** selected_detections = get_actual_detections(dets, num, classes, thresh, &selected_detections_num);

    // text output
    qsort(selected_detections, selected_detections_num, sizeof(*selected_detections), compare_by_lefts);
    for (int i = 0; i < selected_detections_num; ++i) {
        const int best_class = selected_detections[i]->best_class;
        printf("%s: %.0f%%\t(left: %.0f\ttop: %.0f\tw: %0.f\th: %0.f)\n", names[best_class],
            selected_detections[i]->prob[best_class] * 100,
            (selected_detections[i]->bbox.x - selected_detections[i]->bbox.w / 2)*im.w,
            (selected_detections[i]->bbox.y - selected_detections[i]->bbox.h / 2)*im.h,
            selected_detections[i]->bbox.w*im.w, selected_detections[i]->bbox.h*im.h);
        for (int j = 0; j < classes; ++j) {
            if (selected_detections[i]->prob[j] > thresh && j != best_class) {
                printf("%s: %.0f%%\n", names[j], selected_detections[i]->prob[j] * 100);
            }
        }
    }

    // image output
    qsort(selected_detections, selected_detections_num, sizeof(*selected_detections), compare_by_probs);
    for (int i = 0; i < selected_detections_num; ++i) {
            int width = im.h * .006;
            if (width < 1)
                width = 1;
@@ -261,8 +304,8 @@
            }
            */

            //printf("%d %s: %.0f%%\n", i, names[class_id], prob*100);
            int offset = class_id * 123457 % classes;
            //printf("%d %s: %.0f%%\n", i, names[dets[i].best_class], prob*100);
            int offset = selected_detections[i]->best_class * 123457 % classes;
            float red = get_color(2, offset, classes);
            float green = get_color(1, offset, classes);
            float blue = get_color(0, offset, classes);
@@ -273,7 +316,7 @@
            rgb[0] = red;
            rgb[1] = green;
            rgb[2] = blue;
            box b = dets[i].bbox;
            box b = selected_detections[i]->bbox;
            //printf("%f %f %f %f\n", b.x, b.y, b.w, b.h);

            int left = (b.x - b.w / 2.)*im.w;
@@ -294,12 +337,20 @@

            draw_box_width(im, left, top, right, bot, width, red, green, blue);
            if (alphabet) {
                char labelstr[4096] = { 0 };
                strcat(labelstr, names[selected_detections[i]->best_class]);
                for (int j = 0; j < classes; ++j) {
                    if (selected_detections[i]->prob[j] > thresh && j != selected_detections[i]->best_class) {
                        strcat(labelstr, ", ");
                        strcat(labelstr, names[j]);
                    }
                }
                image label = get_label_v3(alphabet, labelstr, (im.h*.03));
                draw_label(im, top + width, left, label, rgb);
                free_image(label);
            }
            if (dets[i].mask) {
                image mask = float_to_image(14, 14, 1, dets[i].mask);
            if (selected_detections[i]->mask) {
                image mask = float_to_image(14, 14, 1, selected_detections[i]->mask);
                image resized_mask = resize_image(mask, b.w*im.w, b.h*im.h);
                image tmask = threshold_image(resized_mask, .5);
                embed_image(tmask, im, left, top);
@@ -307,8 +358,8 @@
                free_image(resized_mask);
                free_image(tmask);
            }
        }
    }
    free(selected_detections);
}

void draw_detections(image im, int num, float thresh, box *boxes, float **probs, char **names, image **alphabet, int classes)