merge with master

README.md

@@ -389,7 +389,7 @@ Then add to your created project:
 
 2. Then stop and by using partially-trained model `/backup/yolov4_1000.weights` run training with multigpu (up to 4 GPUs): `darknet.exe detector train cfg/coco.data cfg/yolov4.cfg /backup/yolov4_1000.weights -gpus 0,1,2,3`
 
-Only for small datasets sometimes better to decrease learning rate, for 4 GPUs set `learning_rate = 0.00025` (i.e. learning_rate = 0.001 / GPUs). In this case also increase 4x times `burn_in =` and `max_batches =` in your cfg-file. I.e. use `burn_in = 4000` instead of `1000`. Same goes for `steps=` if `policy=steps` is set.
+If you get a Nan, then for some datasets better to decrease learning rate, for 4 GPUs set `learning_rate = 0,00065` (i.e. learning_rate = 0.00261 / GPUs). In this case also increase 4x times `burn_in =` in your cfg-file. I.e. use `burn_in = 4000` instead of `1000`.
 
 https://groups.google.com/d/msg/darknet/NbJqonJBTSY/Te5PfIpuCAAJ
 

build/darknet/x64/cfg/yolov4.cfg

@@ -1,10 +1,9 @@
 [net]
-# Testing
-#batch=1
-#subdivisions=1
-# Training
 batch=64
 subdivisions=8
+# Training
+#width=512
+#height=512
 width=608
 height=608
 channels=3

build/darknet/x64/darknet.py

@@ -313,7 +313,7 @@ netMain = None
 metaMain = None
 altNames = None
 
-def performDetect(imagePath="data/dog.jpg", thresh= 0.25, configPath = "./cfg/yolov3.cfg", weightPath = "yolov3.weights", metaPath= "./cfg/coco.data", showImage= True, makeImageOnly = False, initOnly= False):
+def performDetect(imagePath="data/dog.jpg", thresh= 0.25, configPath = "./cfg/yolov4.cfg", weightPath = "yolov4.weights", metaPath= "./cfg/coco.data", showImage= True, makeImageOnly = False, initOnly= False):
     """
     Convenience function to handle the detection and returns of objects.
 
@@ -456,7 +456,7 @@ def performDetect(imagePath="data/dog.jpg", thresh= 0.25, configPath = "./cfg/yo
             print("Unable to show image: "+str(e))
     return detections
 
-def performBatchDetect(thresh= 0.25, configPath = "./cfg/yolov3.cfg", weightPath = "yolov3.weights", metaPath= "./cfg/coco.data", hier_thresh=.5, nms=.45, batch_size=3):
+def performBatchDetect(thresh= 0.25, configPath = "./cfg/yolov4.cfg", weightPath = "yolov4.weights", metaPath= "./cfg/coco.data", hier_thresh=.5, nms=.45, batch_size=3):
     import cv2
     import numpy as np
     # NB! Image sizes should be the same

build/darknet/x64/darknet_video.py

@@ -42,8 +42,8 @@ altNames = None
 def YOLO():
 
     global metaMain, netMain, altNames
-    configPath = "./cfg/yolov3.cfg"
-    weightPath = "./yolov3.weights"
+    configPath = "./cfg/yolov4.cfg"
+    weightPath = "./yolov4.weights"
     metaPath = "./cfg/coco.data"
     if not os.path.exists(configPath):
         raise ValueError("Invalid config path `" +

cfg/yolov4.cfg

@@ -1,10 +1,9 @@
 [net]
-# Testing
-#batch=1
-#subdivisions=1
-# Training
 batch=64
 subdivisions=8
+# Training
+#width=512
+#height=512
 width=608
 height=608
 channels=3

darknet.py

@@ -313,7 +313,7 @@ netMain = None
 metaMain = None
 altNames = None
 
-def performDetect(imagePath="data/dog.jpg", thresh= 0.25, configPath = "./cfg/yolov3.cfg", weightPath = "yolov3.weights", metaPath= "./cfg/coco.data", showImage= True, makeImageOnly = False, initOnly= False):
+def performDetect(imagePath="data/dog.jpg", thresh= 0.25, configPath = "./cfg/yolov4.cfg", weightPath = "yolov4.weights", metaPath= "./cfg/coco.data", showImage= True, makeImageOnly = False, initOnly= False):
     """
     Convenience function to handle the detection and returns of objects.
 
@@ -456,7 +456,7 @@ def performDetect(imagePath="data/dog.jpg", thresh= 0.25, configPath = "./cfg/yo
             print("Unable to show image: "+str(e))
     return detections
 
-def performBatchDetect(thresh= 0.25, configPath = "./cfg/yolov3.cfg", weightPath = "yolov3.weights", metaPath= "./cfg/coco.data", hier_thresh=.5, nms=.45, batch_size=3):
+def performBatchDetect(thresh= 0.25, configPath = "./cfg/yolov4.cfg", weightPath = "yolov4.weights", metaPath= "./cfg/coco.data", hier_thresh=.5, nms=.45, batch_size=3):
     import cv2
     import numpy as np
     # NB! Image sizes should be the same

darknet_video.py

@@ -42,8 +42,8 @@ altNames = None
 def YOLO():
 
     global metaMain, netMain, altNames
-    configPath = "./cfg/yolov3.cfg"
-    weightPath = "./yolov3.weights"
+    configPath = "./cfg/yolov4.cfg"
+    weightPath = "./yolov4.weights"
     metaPath = "./cfg/coco.data"
     if not os.path.exists(configPath):
         raise ValueError("Invalid config path `" +

src/activations.c

@@ -293,6 +293,8 @@ float gradient(float x, ACTIVATION a)
             return loggy_gradient(x);
         case RELU:
             return relu_gradient(x);
+        case RELU6:
+            return relu6_gradient(x);
         case NORM_CHAN:
             //return relu_gradient(x);
         case NORM_CHAN_SOFTMAX_MAXVAL: