How to use a trained params to predict?

[zik1ng]

Hi. I am a new one here. It is a awesome lib for CV, thanks for your great work. Now, I am encountering a problem that I trained ssd_300_vgg16_atrous_voc on my own dataset, the dataset has 4 classes. After training, I got a "ssd_300_vgg16_atrous_voc_best.params" file. When I want to load the parmeters and predict on new images, it return an error, logs as it, AssertionError: Failed loading Parameter 'ssd0_convpredictor4_conv0_weight' from saved params: shape incompatible expected (126, 0, 3, 3) vs saved (30, 512, 3, 3) The command I used as follow, from gluoncv import model_zoo, data, utils from matplotlib import pyplot as plt import mxnet as mx ctx = [mx.cpu()] net = model_zoo.get_model('ssd_300_vgg16_atrous_voc', pretrained=False) net.load_parameters('ssd_300_vgg16_atrous_voc_best.params', ctx = ctx)

Thanks for any help.

[zik1ng]

And could you please add a function draw boxes on image then return this image? I found it is difficult for me to use cv2.imshow() show processed images in real time detection. Thank you.

[zhreshold]

because you are assigning your customized weights to voc models, the shape of class prediction is different. You can load the same net in your training script.

I have added a convenient getter for custom dataset just merged in master

net = gcv.model_zoo.get_model('sad_300_vgg16_atrous_custom', classes=['a','b','c','d'])

You can use it to create a custom network quickly.

[zik1ng]

Thank you for your useful reply :)

[zik1ng]

Thank you again for your reply, that really solved my problem. But there is a new question that I wanna use this trained model make a webcam real-time detection. I use opencv to get the images and call cv2.imshow() to plot images with boxes. How can I get the image matrix with boxes in real time? Thanks. @zhreshold

[ZengyuanYu]

@xDooooot Hi LaoGe, I trained SSD and get some params, when I load it and predict my test image, it is pick a issue, please you upload some code about you?

[zik1ng]

Here is what I used. `

Load model

net = gcv.model_zoo.get_model('yolo3_darknet53_custom', classes=['custom_label1', 'custom_label2', ...], pretrained_base=False)
net.load_parameters('yolo3_darknet53_voc_best.params')

Get images

cap = cv2.VideoCapture(0)

Load frame from the camera

ret, frame = cap.read()

# Image pre-processing
frame = mx.nd.array(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)).astype('uint8')
rgb_nd, frame = gcv.data.transforms.presets.ssd.transform_test(frame, short=512, max_size=700)

# Run frame through network
class_IDs, scores, bounding_boxes = net(rgb_nd)

` 希望能帮到你 :)

[zik1ng]

@ZengyuanYu

[ZengyuanYu]

@xDooooot Thank you! I got it though others ask. But use camera also funny to try it.

[FAFACHR]

Hello, I'm new in computer vision field. I'm using gluoncv to do instance segmentation with mask-rcnn but i have some problems. I did my own dataset similar to coco format and I trained with this model "mask_rcnn_resnet50_v1b_coco", but my dataset has just 2 classes. After training i got the parameter file and when i loaded the parameters and predict a new image, the classification was wrong and still loaded all classes of the coco dataset (80 classes). So is there any solution to customize the number of classes?