Skip to content

Instantly share code, notes, and snippets.

View f-rumblefish's full-sized avatar

f-rumblefish

4 followers · 1 following
View GitHub Profile
@f-rumblefish
f-rumblefish / YOLOv3 Checklist.csv
Last active June 17, 2019 02:21
A Checklist for Training YOLOv3
0. Source https://github.com/qqwweee/keras-yolo3
1. Training 1.1. create mydata_train.txt ref: train.txt
1.2. create mydata_class.txt ref: model_data/voc_classes.txt
1.3. change train.py to train_mydata.py 1.3.1. annotation_path = 'mydata_train.txt' ref: 1.1
1.3.2. classes_path = 'model_data/mydata_classes.txt' ref: 1.2
1.3.3. create_model with load_pretrained=False optional (if pre-trained weight doesn't work)
1.3.4. model.save('mymodel.h5') save model and weight for prediction
1.3.5. change epochs from 50 to 5 optional (check if it works or not)
1.3.6. change batch_size from 32 to 8 optional (if it has some memory problems)
1.3.7. remove training in the second stage optional (if the firs
@f-rumblefish
f-rumblefish / MobileNet_log.txt
Last active February 7, 2019 12:17
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_1 (InputLayer) (None, 224, 224, 3) 0
_________________________________________________________________
conv1_pad (ZeroPadding2D) (None, 225, 225, 3) 0
_________________________________________________________________
conv1 (Conv2D) (None, 112, 112, 32) 864
_________________________________________________________________
conv1_bn (BatchNormalization (None, 112, 112, 32) 128
@f-rumblefish
f-rumblefish / Model_InceptionV3.py
Last active October 7, 2019 01:35
from keras.applications.inception_v3 import InceptionV3
from keras.models import Model
from keras.layers import Dense, GlobalAveragePooling2D
# parameters for architecture
input_shape = (224, 224, 3)
num_classes = 6
conv_size = 32
# parameters for training
@f-rumblefish
f-rumblefish / Model_SimpleNet.py
Created February 7, 2019 01:17
from keras.models import Sequential
from keras.layers import Conv2D, MaxPooling2D, Flatten, Dense
# parameters for architecture
input_shape = (224, 224, 3)
num_classes = 6
conv_size = 32
# parameters for training
batch_size = 32
@f-rumblefish
f-rumblefish / Model_MobileNet.py
Last active November 20, 2019 01:07
from keras.applications.mobilenet import MobileNet
from keras.models import Model
from keras.layers import Dense, GlobalAveragePooling2D
# parameters for architecture
input_shape = (224, 224, 3)
num_classes = 6
conv_size = 32
# parameters for training
@f-rumblefish
f-rumblefish / class_90.py
Created October 23, 2018 15:57
classes_90 = ["background", "person", "bicycle", "car", "motorcycle",
"airplane", "bus", "train", "truck", "boat", "traffic light", "fire hydrant",
"unknown", "stop sign", "parking meter", "bench", "bird", "cat", "dog", "horse",
"sheep", "cow", "elephant", "bear", "zebra", "giraffe", "unknown", "backpack",
"umbrella", "unknown", "unknown", "handbag", "tie", "suitcase", "frisbee", "skis",
"snowboard", "sports ball", "kite", "baseball bat", "baseball glove", "skateboard",
"surfboard", "tennis racket", "bottle", "unknown", "wine glass", "cup", "fork", "knife",
"spoon", "bowl", "banana", "apple", "sandwich", "orange", "broccoli", "carrot", "hot dog",
"pizza", "donut", "cake", "chair", "couch", "potted plant", "bed", "unknown", "dining table",
"unknown", "unknown", "toilet", "unknown", "tv", "laptop", "mouse", "remote", "keyboard",
@f-rumblefish
f-rumblefish / class_80.py
Created October 23, 2018 15:56
classes_80 = ["person", "bicycle", "car", "motorcycle",
"airplane", "bus", "train", "truck", "boat", "traffic light", "fire hydrant",
"stop sign", "parking meter", "bench", "bird", "cat", "dog", "horse",
"sheep", "cow", "elephant", "bear", "zebra", "giraffe", "backpack",
"umbrella", "handbag", "tie", "suitcase", "frisbee", "skis",
"snowboard", "sports ball", "kite", "baseball bat", "baseball glove", "skateboard",
"surfboard", "tennis racket", "bottle", "wine glass", "cup", "fork", "knife",
"spoon", "bowl", "banana", "apple", "sandwich", "orange", "broccoli", "carrot", "hot dog",
"pizza", "donut", "cake", "chair", "couch", "potted plant", "bed", "dining table",
"toilet", "tv", "laptop", "mouse", "remote", "keyboard",
@f-rumblefish
f-rumblefish / yolo_prediction.py
Last active October 23, 2018 15:43
YOLOv2 Prediction
# loop 845 bounding boxes
for i in range(Net_YOLO_pred.shape[0]):
# get the confidence on the object
confidence_on_box = Net_YOLO_pred[i][4]
# find the class index with the highest probability
probability_list = Net_YOLO_pred[i][5:]
class_index = probability_list.argmax(axis=0)
probability_on_class = probability_list[class_index]
# get the score
score = confidence_on_box * probability_on_class
@f-rumblefish
f-rumblefish / ssd_prediction.py
Created October 23, 2018 15:23
SSD/MobileNet Prediction
# loop 100 bounding boxes
for detection in Net_SSD_pred[0,0,:,:]:
# get the class index
class_index = int(detection[1])
# get the score
score = float(detection[2])
# draw the bounding box
if score > threshold:
left = detection[3] * cols
top = detection[4] * rows
@f-rumblefish
f-rumblefish / object_detection_ssd.py
Last active October 23, 2018 01:59
Object Detection via SSD/MobileNet
pb = 'frozen_inference_graph.pb'
pbt = 'ssd_mobilenet_v1_coco_2017_11_17.pbtxt'
Net_SSD = cv.dnn.readNetFromTensorflow(pb,pbt)
Net_SSD.setInput(cv.dnn.blobFromImage(img, 1.0/127.5, (300, 300), (127.5, 127.5, 127.5), swapRB = True, crop = False))
Net_SSD_pred = Net_SSD.forward()
print(">>> SSD/MobileNet prediction shape = ", Net_SSD_pred.shape)