🐍 🐍 🐍 🐍

.gitignore

@@ -4,6 +4,7 @@
 *.out
 *.png
 *.jpg
+*.pyc
 old/
 mnist/
 data/

Makefile

@@ -10,21 +10,21 @@ ARCH= -gencode arch=compute_20,code=[sm_20,sm_21] \
       -gencode arch=compute_52,code=[sm_52,compute_52]
 
 # This is what I use, uncomment if you know your arch and want to specify
-# ARCH=  -gencode arch=compute_52,code=compute_52
+ARCH= -gencode arch=compute_52,code=compute_52
 
 VPATH=./src/:./examples
-LIB=libdarknet.a
+LIB=libdarknet.so
 EXEC=darknet
 OBJDIR=./obj/
 
 CC=gcc
-NVCC=nvcc 
+NVCC=nvcc --compiler-options '-fPIC'
 AR=ar
 ARFLAGS=-rv
 OPTS=-Ofast
 LDFLAGS= -lm -pthread 
 COMMON= -Iinclude/ -Isrc/
-CFLAGS=-Wall -Wfatal-errors 
+CFLAGS=-Wall -Wfatal-errors -fPIC
 
 ifeq ($(DEBUG), 1) 
 OPTS=-O0 -g
@@ -69,7 +69,7 @@ $(EXEC): $(EXECOBJ) $(LIB)
 	$(CC) $(COMMON) $(CFLAGS) $^ -o $@ $(LDFLAGS) $(LIB)
 
 $(LIB): $(OBJS)
-	$(AR) $(ARFLAGS) $@ $^ 
+	$(CC) $(CFLAGS) -shared $^ -o $@ $(LDFLAGS)
 
 $(OBJDIR)%.o: %.c $(DEPS)
 	$(CC) $(COMMON) $(CFLAGS) -c $< -o $@

cfg/cifar.cfg

@@ -1,25 +1,23 @@
 [net]
 batch=128
 subdivisions=1
-height=32
-width=32
+height=28
+width=28
 channels=3
-momentum=0.9
-decay=0.0005
+max_crop=32
+min_crop=32
+
+hue=.1
+saturation=.75
+exposure=.75
 
 learning_rate=0.4
 policy=poly
 power=4
-max_batches = 50000
-
-[crop]
-crop_width=28
-crop_height=28
-flip=1
-angle=0
-saturation = 1
-exposure = 1
-noadjust=1
+max_batches = 5000
+momentum=0.9
+decay=0.0005
+
 
 [convolutional]
 batch_normalize=1

examples/classifier.c

@@ -33,7 +33,6 @@ void train_classifier(char *datacfg, char *cfgfile, char *weightfile, int *gpus,
         cuda_set_device(gpus[i]);
 #endif
         nets[i] = load_network(cfgfile, weightfile, clear);
-        nets[i].learning_rate *= ngpus;
     }
     srand(time(0));
     network net = nets[0];

examples/coco.c

@@ -313,7 +313,7 @@ void test_coco(char *cfgfile, char *weightfile, char *filename, float thresh)
     if(weightfile){
         load_weights(&net, weightfile);
     }
-    detection_layer l = net.layers[net.n-1];
+    layer l = net.layers[net.n-1];
     set_batch_network(&net, 1);
     srand(2222222);
     float nms = .4;

examples/detector.c

@@ -22,12 +22,7 @@ void train_detector(char *datacfg, char *cfgfile, char *weightfile, int *gpus, i
 #ifdef GPU
         cuda_set_device(gpus[i]);
 #endif
-        nets[i] = parse_network_cfg(cfgfile);
-        if(weightfile){
-            load_weights(&nets[i], weightfile);
-        }
-        if(clear) *nets[i].seen = 0;
-        nets[i].learning_rate *= ngpus;
+        nets[i] = load_network(cfgfile, weightfile, clear);
     }
     srand(time(0));
     network net = nets[0];

examples/lsd.c

@@ -451,7 +451,7 @@ void train_dcgan(char *cfg, char *weight, char *acfg, char *aweight, int clear,
     printf("%s\n", base);
     network gnet = load_network(cfg, weight, clear);
     network anet = load_network(acfg, aweight, clear);
-    float orig_rate = anet.learning_rate;
+    //float orig_rate = anet.learning_rate;
 
     int start = 0;
     int i, j, k;
@@ -494,7 +494,7 @@ void train_dcgan(char *cfg, char *weight, char *acfg, char *aweight, int clear,
     int y_size = gnet.truths*gnet.batch;
     float *imerror = cuda_make_array(0, y_size);
 
-    int ay_size = anet.truths*anet.batch;
+    //int ay_size = anet.truths*anet.batch;
 
     float aloss_avg = -1;
 
@@ -664,14 +664,14 @@ void train_colorizer(char *cfg, char *weight, char *acfg, char *aweight, int cle
     clock_t time;
 
     int x_size = net.inputs*net.batch;
-    int y_size = x_size;
+    //int y_size = x_size;
     net.delta = 0;
     net.train = 1;
     float *pixs = calloc(x_size, sizeof(float));
     float *graypixs = calloc(x_size, sizeof(float));
-    float *y = calloc(y_size, sizeof(float));
+    //float *y = calloc(y_size, sizeof(float));
 
-    int ay_size = anet.outputs*anet.batch;
+    //int ay_size = anet.outputs*anet.batch;
     anet.delta = 0;
     anet.train = 1;
 

examples/nightmare.c

@@ -1,5 +1,7 @@
 #include "darknet.h"
 
+#include <math.h>
+
 // ./darknet nightmare cfg/extractor.recon.cfg ~/trained/yolo-coco.conv frame6.png -reconstruct -iters 500 -i 3 -lambda .1 -rate .01 -smooth 2
 
 float abs_mean(float *x, int n)
@@ -128,11 +130,11 @@ void smooth(image recon, image update, float lambda, int num)
 void reconstruct_picture(network net, float *features, image recon, image update, float rate, float momentum, float lambda, int smooth_size, int iters)
 {
     int iter = 0;
-    layer l = get_network_output_layer(net);
     for (iter = 0; iter < iters; ++iter) {
         image delta = make_image(recon.w, recon.h, recon.c);
 
 #ifdef GPU
+        layer l = get_network_output_layer(net);
         cuda_push_array(net.input_gpu, recon.data, recon.w*recon.h*recon.c);
         //cuda_push_array(net.truth_gpu, features, net.truths);
         net.delta_gpu = cuda_make_array(delta.data, delta.w*delta.h*delta.c);

examples/rnn.c

@@ -1,5 +1,7 @@
 #include "darknet.h"
 
+#include <math.h>
+
 typedef struct {
     float *x;
     float *y;

examples/segmenter.c

@@ -24,7 +24,6 @@ void train_segmenter(char *datacfg, char *cfgfile, char *weightfile, int *gpus,
             load_weights(&nets[i], weightfile);
         }
         if(clear) *nets[i].seen = 0;
-        nets[i].learning_rate *= ngpus;
     }
     srand(time(0));
     network net = nets[0];
@@ -76,6 +75,15 @@ void train_segmenter(char *datacfg, char *cfgfile, char *weightfile, int *gpus,
         pthread_join(load_thread, 0);
         train = buffer;
         load_thread = load_data(args);
+        image tr = float_to_image(net.w, net.h, 81, train.y.vals[0]);
+        image im = float_to_image(net.w, net.h, net.c, train.X.vals[0]);
+        image mask = mask_to_rgb(tr);
+        show_image(im, "input");
+        show_image(mask, "truth");
+#ifdef OPENCV
+        cvWaitKey(100);
+#endif
+        free_image(mask);
 
         printf("Loaded: %lf seconds\n", sec(clock()-time));
         time=clock();

examples/yolo.c

@@ -279,7 +279,7 @@ void test_yolo(char *cfgfile, char *weightfile, char *filename, float thresh)
     if(weightfile){
         load_weights(&net, weightfile);
     }
-    detection_layer l = net.layers[net.n-1];
+    layer l = net.layers[net.n-1];
     set_batch_network(&net, 1);
     srand(2222222);
     clock_t time;