up new v4

Scheaven

2021-06-03 168af40fe9a3cc81c6ee16b3e81f154780c36bdb

 lib/detecter_tools/darknet/detection_layer.c

@@ -1,315 +1,315 @@
#include "detection_layer.h"

#include "activations.h"

#include "softmax_layer.h"

#include "blas.h"

#include "box.h"

#include "dark_cuda.h"

#include "utils.h"

#include <stdio.h>

#include <assert.h>

#include <string.h>

#include <stdlib.h>



detection_layer make_detection_layer(int batch, int inputs, int n, int side, int classes, int coords, int rescore)

{

    detection_layer l = { (LAYER_TYPE)0 };

    l.type = DETECTION;



    l.n = n;

    l.batch = batch;

    l.inputs = inputs;

    l.classes = classes;

    l.coords = coords;

    l.rescore = rescore;

    l.side = side;

    l.w = side;

    l.h = side;

    assert(side*side*((1 + l.coords)*l.n + l.classes) == inputs);

    l.cost = (float*)xcalloc(1, sizeof(float));

    l.outputs = l.inputs;

    l.truths = l.side*l.side*(1+l.coords+l.classes);

    l.output = (float*)xcalloc(batch * l.outputs, sizeof(float));

    l.delta = (float*)xcalloc(batch * l.outputs, sizeof(float));



    l.forward = forward_detection_layer;

    l.backward = backward_detection_layer;

#ifdef GPU

    l.forward_gpu = forward_detection_layer_gpu;

    l.backward_gpu = backward_detection_layer_gpu;

    l.output_gpu = cuda_make_array(l.output, batch*l.outputs);

    l.delta_gpu = cuda_make_array(l.delta, batch*l.outputs);

#endif



    fprintf(stderr, "Detection Layer\n");

    srand(time(0));



    return l;

}



void forward_detection_layer(const detection_layer l, network_state state)

{

    int locations = l.side*l.side;

    int i,j;

    memcpy(l.output, state.input, l.outputs*l.batch*sizeof(float));

    //if(l.reorg) reorg(l.output, l.w*l.h, size*l.n, l.batch, 1);

    int b;

    if (l.softmax){

        for(b = 0; b < l.batch; ++b){

            int index = b*l.inputs;

            for (i = 0; i < locations; ++i) {

                int offset = i*l.classes;

                softmax(l.output + index + offset, l.classes, 1,

                        l.output + index + offset, 1);

            }

        }

    }

    if(state.train){

        float avg_iou = 0;

        float avg_cat = 0;

        float avg_allcat = 0;

        float avg_obj = 0;

        float avg_anyobj = 0;

        int count = 0;

        *(l.cost) = 0;

        int size = l.inputs * l.batch;

        memset(l.delta, 0, size * sizeof(float));

        for (b = 0; b < l.batch; ++b){

            int index = b*l.inputs;

            for (i = 0; i < locations; ++i) {

                int truth_index = (b*locations + i)*(1+l.coords+l.classes);

                int is_obj = state.truth[truth_index];

                for (j = 0; j < l.n; ++j) {

                    int p_index = index + locations*l.classes + i*l.n + j;

                    l.delta[p_index] = l.noobject_scale*(0 - l.output[p_index]);

                    *(l.cost) += l.noobject_scale*pow(l.output[p_index], 2);

                    avg_anyobj += l.output[p_index];

                }



                int best_index = -1;

                float best_iou = 0;

                float best_rmse = 20;



                if (!is_obj){

                    continue;

                }



                int class_index = index + i*l.classes;

                for(j = 0; j < l.classes; ++j) {

                    l.delta[class_index+j] = l.class_scale * (state.truth[truth_index+1+j] - l.output[class_index+j]);

                    *(l.cost) += l.class_scale * pow(state.truth[truth_index+1+j] - l.output[class_index+j], 2);

                    if(state.truth[truth_index + 1 + j]) avg_cat += l.output[class_index+j];

                    avg_allcat += l.output[class_index+j];

                }



                box truth = float_to_box(state.truth + truth_index + 1 + l.classes);

                truth.x /= l.side;

                truth.y /= l.side;



                for(j = 0; j < l.n; ++j){

                    int box_index = index + locations*(l.classes + l.n) + (i*l.n + j) * l.coords;

                    box out = float_to_box(l.output + box_index);

                    out.x /= l.side;

                    out.y /= l.side;



                    if (l.sqrt){

                        out.w = out.w*out.w;

                        out.h = out.h*out.h;

                    }



                    float iou  = box_iou(out, truth);

                    //iou = 0;

                    float rmse = box_rmse(out, truth);

                    if(best_iou > 0 || iou > 0){

                        if(iou > best_iou){

                            best_iou = iou;

                            best_index = j;

                        }

                    }else{

                        if(rmse < best_rmse){

                            best_rmse = rmse;

                            best_index = j;

                        }

                    }

                }



                if(l.forced){

                    if(truth.w*truth.h < .1){

                        best_index = 1;

                    }else{

                        best_index = 0;

                    }

                }

                if(l.random && *(state.net.seen) < 64000){

                    best_index = rand()%l.n;

                }



                int box_index = index + locations*(l.classes + l.n) + (i*l.n + best_index) * l.coords;

                int tbox_index = truth_index + 1 + l.classes;



                box out = float_to_box(l.output + box_index);

                out.x /= l.side;

                out.y /= l.side;

                if (l.sqrt) {

                    out.w = out.w*out.w;

                    out.h = out.h*out.h;

                }

                float iou  = box_iou(out, truth);



                //printf("%d,", best_index);

                int p_index = index + locations*l.classes + i*l.n + best_index;

                *(l.cost) -= l.noobject_scale * pow(l.output[p_index], 2);

                *(l.cost) += l.object_scale * pow(1-l.output[p_index], 2);

                avg_obj += l.output[p_index];

                l.delta[p_index] = l.object_scale * (1.-l.output[p_index]);



                if(l.rescore){

                    l.delta[p_index] = l.object_scale * (iou - l.output[p_index]);

                }



                l.delta[box_index+0] = l.coord_scale*(state.truth[tbox_index + 0] - l.output[box_index + 0]);

                l.delta[box_index+1] = l.coord_scale*(state.truth[tbox_index + 1] - l.output[box_index + 1]);

                l.delta[box_index+2] = l.coord_scale*(state.truth[tbox_index + 2] - l.output[box_index + 2]);

                l.delta[box_index+3] = l.coord_scale*(state.truth[tbox_index + 3] - l.output[box_index + 3]);

                if(l.sqrt){

                    l.delta[box_index+2] = l.coord_scale*(sqrt(state.truth[tbox_index + 2]) - l.output[box_index + 2]);

                    l.delta[box_index+3] = l.coord_scale*(sqrt(state.truth[tbox_index + 3]) - l.output[box_index + 3]);

                }



                *(l.cost) += pow(1-iou, 2);

                avg_iou += iou;

                ++count;

            }

        }



        if(0){

            float* costs = (float*)xcalloc(l.batch * locations * l.n, sizeof(float));

            for (b = 0; b < l.batch; ++b) {

                int index = b*l.inputs;

                for (i = 0; i < locations; ++i) {

                    for (j = 0; j < l.n; ++j) {

                        int p_index = index + locations*l.classes + i*l.n + j;

                        costs[b*locations*l.n + i*l.n + j] = l.delta[p_index]*l.delta[p_index];

                    }

                }

            }

            int indexes[100];

            top_k(costs, l.batch*locations*l.n, 100, indexes);

            float cutoff = costs[indexes[99]];

            for (b = 0; b < l.batch; ++b) {

                int index = b*l.inputs;

                for (i = 0; i < locations; ++i) {

                    for (j = 0; j < l.n; ++j) {

                        int p_index = index + locations*l.classes + i*l.n + j;

                        if (l.delta[p_index]*l.delta[p_index] < cutoff) l.delta[p_index] = 0;

                    }

                }

            }

            free(costs);

        }





        *(l.cost) = pow(mag_array(l.delta, l.outputs * l.batch), 2);





        printf("Detection Avg IOU: %f, Pos Cat: %f, All Cat: %f, Pos Obj: %f, Any Obj: %f, count: %d\n", avg_iou/count, avg_cat/count, avg_allcat/(count*l.classes), avg_obj/count, avg_anyobj/(l.batch*locations*l.n), count);

        //if(l.reorg) reorg(l.delta, l.w*l.h, size*l.n, l.batch, 0);

    }

}



void backward_detection_layer(const detection_layer l, network_state state)

{

    axpy_cpu(l.batch*l.inputs, 1, l.delta, 1, state.delta, 1);

}



void get_detection_boxes(layer l, int w, int h, float thresh, float **probs, box *boxes, int only_objectness)

{

    int i,j,n;

    float *predictions = l.output;

    //int per_cell = 5*num+classes;

    for (i = 0; i < l.side*l.side; ++i){

        int row = i / l.side;

        int col = i % l.side;

        for(n = 0; n < l.n; ++n){

            int index = i*l.n + n;

            int p_index = l.side*l.side*l.classes + i*l.n + n;

            float scale = predictions[p_index];

            int box_index = l.side*l.side*(l.classes + l.n) + (i*l.n + n)*4;

            boxes[index].x = (predictions[box_index + 0] + col) / l.side * w;

            boxes[index].y = (predictions[box_index + 1] + row) / l.side * h;

            boxes[index].w = pow(predictions[box_index + 2], (l.sqrt?2:1)) * w;

            boxes[index].h = pow(predictions[box_index + 3], (l.sqrt?2:1)) * h;

            for(j = 0; j < l.classes; ++j){

                int class_index = i*l.classes;

                float prob = scale*predictions[class_index+j];

                probs[index][j] = (prob > thresh) ? prob : 0;

            }

            if(only_objectness){

                probs[index][0] = scale;

            }

        }

    }

}



#ifdef GPU



void forward_detection_layer_gpu(const detection_layer l, network_state state)

{

    if(!state.train){

        copy_ongpu(l.batch*l.inputs, state.input, 1, l.output_gpu, 1);

        return;

    }



    float* in_cpu = (float*)xcalloc(l.batch * l.inputs, sizeof(float));

    float *truth_cpu = 0;

    if(state.truth){

        int num_truth = l.batch*l.side*l.side*(1+l.coords+l.classes);

        truth_cpu = (float*)xcalloc(num_truth, sizeof(float));

        cuda_pull_array(state.truth, truth_cpu, num_truth);

    }

    cuda_pull_array(state.input, in_cpu, l.batch*l.inputs);

    network_state cpu_state = state;

    cpu_state.train = state.train;

    cpu_state.truth = truth_cpu;

    cpu_state.input = in_cpu;

    forward_detection_layer(l, cpu_state);

    cuda_push_array(l.output_gpu, l.output, l.batch*l.outputs);

    cuda_push_array(l.delta_gpu, l.delta, l.batch*l.inputs);

    free(cpu_state.input);

    if(cpu_state.truth) free(cpu_state.truth);

}



void backward_detection_layer_gpu(detection_layer l, network_state state)

{

    axpy_ongpu(l.batch*l.inputs, 1, l.delta_gpu, 1, state.delta, 1);

    //copy_ongpu(l.batch*l.inputs, l.delta_gpu, 1, state.delta, 1);

}

#endif



void get_detection_detections(layer l, int w, int h, float thresh, detection *dets)

{

   int i, j, n;

   float *predictions = l.output;

   //int per_cell = 5*num+classes;

   for (i = 0; i < l.side*l.side; ++i) {

      int row = i / l.side;

      int col = i % l.side;

      for (n = 0; n < l.n; ++n) {

         int index = i*l.n + n;

         int p_index = l.side*l.side*l.classes + i*l.n + n;

         float scale = predictions[p_index];

         int box_index = l.side*l.side*(l.classes + l.n) + (i*l.n + n) * 4;

         box b;

         b.x = (predictions[box_index + 0] + col) / l.side * w;

         b.y = (predictions[box_index + 1] + row) / l.side * h;

         b.w = pow(predictions[box_index + 2], (l.sqrt ? 2 : 1)) * w;

         b.h = pow(predictions[box_index + 3], (l.sqrt ? 2 : 1)) * h;

         dets[index].bbox = b;

         dets[index].objectness = scale;

         for (j = 0; j < l.classes; ++j) {

            int class_index = i*l.classes;

            float prob = scale*predictions[class_index + j];

            dets[index].prob[j] = (prob > thresh) ? prob : 0;

         }

      }

   }

}

#include "detection_layer.h"
#include "activations.h"
#include "softmax_layer.h"
#include "blas.h"
#include "box.h"
#include "dark_cuda.h"
#include "utils.h"
#include <stdio.h>
#include <assert.h>
#include <string.h>
#include <stdlib.h>

detection_layer make_detection_layer(int batch, int inputs, int n, int side, int classes, int coords, int rescore)
{
    detection_layer l = { (LAYER_TYPE)0 };
    l.type = DETECTION;

    l.n = n;
    l.batch = batch;
    l.inputs = inputs;
    l.classes = classes;
    l.coords = coords;
    l.rescore = rescore;
    l.side = side;
    l.w = side;
    l.h = side;
    assert(side*side*((1 + l.coords)*l.n + l.classes) == inputs);
    l.cost = (float*)xcalloc(1, sizeof(float));
    l.outputs = l.inputs;
    l.truths = l.side*l.side*(1+l.coords+l.classes);
    l.output = (float*)xcalloc(batch * l.outputs, sizeof(float));
    l.delta = (float*)xcalloc(batch * l.outputs, sizeof(float));

    l.forward = forward_detection_layer;
    l.backward = backward_detection_layer;
#ifdef GPU
    l.forward_gpu = forward_detection_layer_gpu;
    l.backward_gpu = backward_detection_layer_gpu;
    l.output_gpu = cuda_make_array(l.output, batch*l.outputs);
    l.delta_gpu = cuda_make_array(l.delta, batch*l.outputs);
#endif

    fprintf(stderr, "Detection Layer\n");
    srand(time(0));

    return l;
}

void forward_detection_layer(const detection_layer l, network_state state)
{
    int locations = l.side*l.side;
    int i,j;
    memcpy(l.output, state.input, l.outputs*l.batch*sizeof(float));
    //if(l.reorg) reorg(l.output, l.w*l.h, size*l.n, l.batch, 1);
    int b;
    if (l.softmax){
        for(b = 0; b < l.batch; ++b){
            int index = b*l.inputs;
            for (i = 0; i < locations; ++i) {
                int offset = i*l.classes;
                softmax(l.output + index + offset, l.classes, 1,
                        l.output + index + offset, 1);
            }
        }
    }
    if(state.train){
        float avg_iou = 0;
        float avg_cat = 0;
        float avg_allcat = 0;
        float avg_obj = 0;
        float avg_anyobj = 0;
        int count = 0;
        *(l.cost) = 0;
        int size = l.inputs * l.batch;
        memset(l.delta, 0, size * sizeof(float));
        for (b = 0; b < l.batch; ++b){
            int index = b*l.inputs;
            for (i = 0; i < locations; ++i) {
                int truth_index = (b*locations + i)*(1+l.coords+l.classes);
                int is_obj = state.truth[truth_index];
                for (j = 0; j < l.n; ++j) {
                    int p_index = index + locations*l.classes + i*l.n + j;
                    l.delta[p_index] = l.noobject_scale*(0 - l.output[p_index]);
                    *(l.cost) += l.noobject_scale*pow(l.output[p_index], 2);
                    avg_anyobj += l.output[p_index];
                }

                int best_index = -1;
                float best_iou = 0;
                float best_rmse = 20;

                if (!is_obj){
                    continue;
                }

                int class_index = index + i*l.classes;
                for(j = 0; j < l.classes; ++j) {
                    l.delta[class_index+j] = l.class_scale * (state.truth[truth_index+1+j] - l.output[class_index+j]);
                    *(l.cost) += l.class_scale * pow(state.truth[truth_index+1+j] - l.output[class_index+j], 2);
                    if(state.truth[truth_index + 1 + j]) avg_cat += l.output[class_index+j];
                    avg_allcat += l.output[class_index+j];
                }

                box truth = float_to_box(state.truth + truth_index + 1 + l.classes);
                truth.x /= l.side;
                truth.y /= l.side;

                for(j = 0; j < l.n; ++j){
                    int box_index = index + locations*(l.classes + l.n) + (i*l.n + j) * l.coords;
                    box out = float_to_box(l.output + box_index);
                    out.x /= l.side;
                    out.y /= l.side;

                    if (l.sqrt){
                        out.w = out.w*out.w;
                        out.h = out.h*out.h;
                    }

                    float iou  = box_iou(out, truth);
                    //iou = 0;
                    float rmse = box_rmse(out, truth);
                    if(best_iou > 0 || iou > 0){
                        if(iou > best_iou){
                            best_iou = iou;
                            best_index = j;
                        }
                    }else{
                        if(rmse < best_rmse){
                            best_rmse = rmse;
                            best_index = j;
                        }
                    }
                }

                if(l.forced){
                    if(truth.w*truth.h < .1){
                        best_index = 1;
                    }else{
                        best_index = 0;
                    }
                }
                if(l.random && *(state.net.seen) < 64000){
                    best_index = rand()%l.n;
                }

                int box_index = index + locations*(l.classes + l.n) + (i*l.n + best_index) * l.coords;
                int tbox_index = truth_index + 1 + l.classes;

                box out = float_to_box(l.output + box_index);
                out.x /= l.side;
                out.y /= l.side;
                if (l.sqrt) {
                    out.w = out.w*out.w;
                    out.h = out.h*out.h;
                }
                float iou  = box_iou(out, truth);

                //printf("%d,", best_index);
                int p_index = index + locations*l.classes + i*l.n + best_index;
                *(l.cost) -= l.noobject_scale * pow(l.output[p_index], 2);
                *(l.cost) += l.object_scale * pow(1-l.output[p_index], 2);
                avg_obj += l.output[p_index];
                l.delta[p_index] = l.object_scale * (1.-l.output[p_index]);

                if(l.rescore){
                    l.delta[p_index] = l.object_scale * (iou - l.output[p_index]);
                }

                l.delta[box_index+0] = l.coord_scale*(state.truth[tbox_index + 0] - l.output[box_index + 0]);
                l.delta[box_index+1] = l.coord_scale*(state.truth[tbox_index + 1] - l.output[box_index + 1]);
                l.delta[box_index+2] = l.coord_scale*(state.truth[tbox_index + 2] - l.output[box_index + 2]);
                l.delta[box_index+3] = l.coord_scale*(state.truth[tbox_index + 3] - l.output[box_index + 3]);
                if(l.sqrt){
                    l.delta[box_index+2] = l.coord_scale*(sqrt(state.truth[tbox_index + 2]) - l.output[box_index + 2]);
                    l.delta[box_index+3] = l.coord_scale*(sqrt(state.truth[tbox_index + 3]) - l.output[box_index + 3]);
                }

                *(l.cost) += pow(1-iou, 2);
                avg_iou += iou;
                ++count;
            }
        }

        if(0){
            float* costs = (float*)xcalloc(l.batch * locations * l.n, sizeof(float));
            for (b = 0; b < l.batch; ++b) {
                int index = b*l.inputs;
                for (i = 0; i < locations; ++i) {
                    for (j = 0; j < l.n; ++j) {
                        int p_index = index + locations*l.classes + i*l.n + j;
                        costs[b*locations*l.n + i*l.n + j] = l.delta[p_index]*l.delta[p_index];
                    }
                }
            }
            int indexes[100];
            top_k(costs, l.batch*locations*l.n, 100, indexes);
            float cutoff = costs[indexes[99]];
            for (b = 0; b < l.batch; ++b) {
                int index = b*l.inputs;
                for (i = 0; i < locations; ++i) {
                    for (j = 0; j < l.n; ++j) {
                        int p_index = index + locations*l.classes + i*l.n + j;
                        if (l.delta[p_index]*l.delta[p_index] < cutoff) l.delta[p_index] = 0;
                    }
                }
            }
            free(costs);
        }


        *(l.cost) = pow(mag_array(l.delta, l.outputs * l.batch), 2);


        printf("Detection Avg IOU: %f, Pos Cat: %f, All Cat: %f, Pos Obj: %f, Any Obj: %f, count: %d\n", avg_iou/count, avg_cat/count, avg_allcat/(count*l.classes), avg_obj/count, avg_anyobj/(l.batch*locations*l.n), count);
        //if(l.reorg) reorg(l.delta, l.w*l.h, size*l.n, l.batch, 0);
    }
}

void backward_detection_layer(const detection_layer l, network_state state)
{
    axpy_cpu(l.batch*l.inputs, 1, l.delta, 1, state.delta, 1);
}

void get_detection_boxes(layer l, int w, int h, float thresh, float **probs, box *boxes, int only_objectness)
{
    int i,j,n;
    float *predictions = l.output;
    //int per_cell = 5*num+classes;
    for (i = 0; i < l.side*l.side; ++i){
        int row = i / l.side;
        int col = i % l.side;
        for(n = 0; n < l.n; ++n){
            int index = i*l.n + n;
            int p_index = l.side*l.side*l.classes + i*l.n + n;
            float scale = predictions[p_index];
            int box_index = l.side*l.side*(l.classes + l.n) + (i*l.n + n)*4;
            boxes[index].x = (predictions[box_index + 0] + col) / l.side * w;
            boxes[index].y = (predictions[box_index + 1] + row) / l.side * h;
            boxes[index].w = pow(predictions[box_index + 2], (l.sqrt?2:1)) * w;
            boxes[index].h = pow(predictions[box_index + 3], (l.sqrt?2:1)) * h;
            for(j = 0; j < l.classes; ++j){
                int class_index = i*l.classes;
                float prob = scale*predictions[class_index+j];
                probs[index][j] = (prob > thresh) ? prob : 0;
            }
            if(only_objectness){
                probs[index][0] = scale;
            }
        }
    }
}

#ifdef GPU

void forward_detection_layer_gpu(const detection_layer l, network_state state)
{
    if(!state.train){
        copy_ongpu(l.batch*l.inputs, state.input, 1, l.output_gpu, 1);
        return;
    }

    float* in_cpu = (float*)xcalloc(l.batch * l.inputs, sizeof(float));
    float *truth_cpu = 0;
    if(state.truth){
        int num_truth = l.batch*l.side*l.side*(1+l.coords+l.classes);
        truth_cpu = (float*)xcalloc(num_truth, sizeof(float));
        cuda_pull_array(state.truth, truth_cpu, num_truth);
    }
    cuda_pull_array(state.input, in_cpu, l.batch*l.inputs);
    network_state cpu_state = state;
    cpu_state.train = state.train;
    cpu_state.truth = truth_cpu;
    cpu_state.input = in_cpu;
    forward_detection_layer(l, cpu_state);
    cuda_push_array(l.output_gpu, l.output, l.batch*l.outputs);
    cuda_push_array(l.delta_gpu, l.delta, l.batch*l.inputs);
    free(cpu_state.input);
    if(cpu_state.truth) free(cpu_state.truth);
}

void backward_detection_layer_gpu(detection_layer l, network_state state)
{
    axpy_ongpu(l.batch*l.inputs, 1, l.delta_gpu, 1, state.delta, 1);
    //copy_ongpu(l.batch*l.inputs, l.delta_gpu, 1, state.delta, 1);
}
#endif

void get_detection_detections(layer l, int w, int h, float thresh, detection *dets)
{
   int i, j, n;
   float *predictions = l.output;
   //int per_cell = 5*num+classes;
   for (i = 0; i < l.side*l.side; ++i) {
      int row = i / l.side;
      int col = i % l.side;
      for (n = 0; n < l.n; ++n) {
         int index = i*l.n + n;
         int p_index = l.side*l.side*l.classes + i*l.n + n;
         float scale = predictions[p_index];
         int box_index = l.side*l.side*(l.classes + l.n) + (i*l.n + n) * 4;
         box b;
         b.x = (predictions[box_index + 0] + col) / l.side * w;
         b.y = (predictions[box_index + 1] + row) / l.side * h;
         b.w = pow(predictions[box_index + 2], (l.sqrt ? 2 : 1)) * w;
         b.h = pow(predictions[box_index + 3], (l.sqrt ? 2 : 1)) * h;
         dets[index].bbox = b;
         dets[index].objectness = scale;
         for (j = 0; j < l.classes; ++j) {
            int class_index = i*l.classes;
            float prob = scale*predictions[class_index + j];
            dets[index].prob[j] = (prob > thresh) ? prob : 0;
         }
      }
   }
}