In [1]:

    

import matplotlib.pyplot as plt
import matplotlib.patches as patches
import numpy as np
import mxnet as mx 
from symbol import get_resnet_model
from symbol import YOLO_loss
from data_ulti import get_iterator
import cv2


    

In [2]:

    

def decodeBox(yolobox, size, dscale):
    i, j, cx, cy, w, h = yolobox
    cxt = j*dscale + cx*dscale
    cyt = i*dscale + cy*dscale
    wt = w*size
    ht = h*size
    return [cxt, cyt, wt, ht]

def bboxdraw(img, label, dscale=32):
    size = img.shape[1]
    ilist, jlist = np.where(label[:,:,0]>0.1)
    
    # Create figure and axes
    fig,ax = plt.subplots(1)
    ax.imshow(np.uint8(img))
    for i,j in zip(ilist, jlist): 
        cx,cy,w,h = label[i,j,1:]
        cxt, cyt, wt ,ht = decodeBox([i, j, cx,cy,w,h], size, dscale)
        # Create a Rectangle patch
        rect = patches.Rectangle((cxt-wt/2,cyt-ht/2), wt,ht,linewidth=1,edgecolor='r',facecolor='none')

        # Add the patch to the Axes
        ax.add_patch(rect)
    
        plt.plot(int(cxt), int(cyt), '*')
    plt.show()


    

In [3]:

    

cat = plt.imread("cat3.jpg")
plt.imshow(cat)
plt.show()
print "Orignal Cat"
W = 224
H = 224
cat_resize = cv2.resize(cat, (H,W))
plt.imshow(cat_resize)
plt.show()
print "Resize Cat"


    

    













    




Orignal Cat






    













    




Resize Cat

In [4]:

    

cat_nd = mx.nd.array(ctx=mx.gpu(0), source_array=cat_resize.transpose((2,0,1)).reshape(1,3,H,W))
cat_itr = mx.io.NDArrayIter(data=cat_nd, data_name='data',  batch_size=1)


    

In [5]:

    

# get sym 
sym, args_params, aux_params = mx.model.load_checkpoint('models/cat_detect_full_scale', 600)
logit = sym.get_internals()['logit_output']
mod = mx.mod.Module(symbol=logit, context=mx.gpu(0))
mod.bind(cat_itr.provide_data)
mod.init_params(allow_missing=False, arg_params=args_params, aux_params=aux_params, 
                initializer=mx.init.Xavier(magnitude=2,rnd_type='gaussian',factor_type='in'))
out = mod.predict(eval_data=cat_itr, num_batch=10)


    

    




/home/chuck/DLenv/MXNet/local/lib/python2.7/site-packages/mxnet/module/base_module.py:52: UserWarning: You created Module with Module(..., label_names=['softmax_label']) but input with name 'softmax_label' is not found in symbol.list_arguments(). Did you mean one of:
	data
  warnings.warn(msg)
/home/chuck/DLenv/MXNet/local/lib/python2.7/site-packages/mxnet/module/base_module.py:64: UserWarning: Data provided by label_shapes don't match names specified by label_names ([] vs. ['softmax_label'])
  warnings.warn(msg)

In [6]:

    

pred->softsing->x/(1+abs(x))
pred = (out.asnumpy()[0]+1)/2
print pred.shape


    

    




(7, 7, 5)

In [7]:

    

print "Prediction"
bboxdraw(cat_resize, pred)


    

    




Prediction