@@ -210,7 +210,7 @@
},
{
"cell_type" : " code"
"execution_count" : 8 ,
"execution_count" : 219 ,
"metadata" : {
"collapsed" : false
},
@@ -258,12 +258,10 @@
" h = F.leaky_relu(self.fc1(h), slope=0.1)\n "
" h = F.dropout(h, train=self.train, ratio=0.5)\n "
" h = F.leaky_relu(self.fc2(h), slope=0.1)\n "
" h = self.fc3(h) # (batch_size, ((5*N_BOXES)+N_CLASSES)*N_GRID*N_GRID)\n "
" \n "
" # extract result tensors\n "
" h = self.fc3(h)\n "
" # reshape predicted tensor\n "
" h = F.reshape(h, (batch_size, (5*N_BOXES)+N_CLASSES, N_GRID, N_GRID))\n "
" x, y, w, h, conf, prob = F.split_axis(h, indices_or_sections=(1,2,3,4,5), axis=1)\n "
" return F.sigmoid(x), F.sigmoid(y), F.sigmoid(w), F.sigmoid(h), F.sigmoid(conf), F.sigmoid(prob)\n "
" return h\n "
" \n "
" def __call__(self, x, t):\n "
" # 推論を実行\n "
@@ -870,7 +868,7 @@
},
{
"cell_type" : " code"
"execution_count" : 96 ,
"execution_count" : 220 ,
"metadata" : {
"collapsed" : false
},
@@ -888,6 +886,215 @@
" ## 推論を実行"
]
},
{
"cell_type" : " code"
"execution_count" : 221 ,
"metadata" : {
"collapsed" : true
},
"outputs" : [],
"source" : [
" h = detector_model.forward(x)"
]
},
{
"cell_type" : " code"
"execution_count" : 261 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [],
"source" : [
" h2 = F.sigmoid(h)"
]
},
{
"cell_type" : " code"
"execution_count" : 234 ,
"metadata" : {
"collapsed" : true
},
"outputs" : [],
"source" : [
" _px, _py, _pw, _ph, _pconf, _pprob = F.split_axis(h.data, indices_or_sections=(1,2,3,4,5), axis=1)\n "
" _px = F.sigmoid(_px).data\n "
" _py = F.sigmoid(_py).data\n "
" _pw = F.sigmoid(_pw).data\n "
" _ph = F.sigmoid(_ph).data\n "
" _pconf = F.sigmoid(_pconf).data\n "
" _pprob = F.sigmoid(_pprob).data"
]
},
{
"cell_type" : " code"
"execution_count" : 267 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [],
"source" : [
" _px2, _py2, _pw2, _ph2, _pconf2, _pprob2 = np.array_split(h2.data, indices_or_sections=(1,2,3,4,5), axis=1)"
]
},
{
"cell_type" : " code"
"execution_count" : 277 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [
{
"data" : {
"text/plain" : [
" array([[-1. , 0.13344583, 0.91010243, 0.1460737 , 0.40128991,\n "
" 0.51702583, 0.57517201],\n "
" [ 0.04876372, 0.29020485, 0.99930859, 0.93551981, 0.05357945,\n "
" 0.86496329, 0.99785376],\n "
" [ 0.82082754, 0.27134973, 0.2106505 , 0.42414856, 0.97827983,\n "
" 0.52681172, 0.92448193],\n "
" [ 0.74471402, 0.22758949, 0.07767206, 0.99378127, 0.43946666,\n "
" 0.63642144, 0.89852059],\n "
" [ 0.11715493, 0.37505376, 0.03614116, 0.00101459, 0.74491358,\n "
" 0.23284879, 0.54585719],\n "
" [ 0.17654884, 0.34766054, 0.34777841, 0.18494397, 0.03876951,\n "
" 0.955625 , 0.42072499],\n "
" [ 0.65734005, 0.7649737 , 0.98942077, 0.39563155, 0.43352816,\n "
" 0.07697165, 0.74792385]], dtype=float32)"
]
},
"execution_count" : 277 ,
"metadata" : {},
"output_type" : " execute_result"
}
],
"source" : [
" _px2[0][0][0][0] = -1\n "
" _px2[0][0]"
]
},
{
"cell_type" : " code"
"execution_count" : 275 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [
{
"data" : {
"text/plain" : [
" (10, 31, 7, 7)"
]
},
"execution_count" : 275 ,
"metadata" : {},
"output_type" : " execute_result"
}
],
"source" : [
" h3 = chainer.cuda.to_cpu(h2.data.copy())\n "
" h3.shape"
]
},
{
"cell_type" : " code"
"execution_count" : 276 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [
{
"data" : {
"text/plain" : [
" array([[ -2.00000000e+00, 1.33445829e-01, 9.10102427e-01,\n "
" 1.46073699e-01, 4.01289910e-01, 5.17025828e-01,\n "
" 5.75172007e-01],\n "
" [ 4.87637222e-02, 2.90204853e-01, 9.99308586e-01,\n "
" 9.35519814e-01, 5.35794497e-02, 8.64963293e-01,\n "
" 9.97853756e-01],\n "
" [ 8.20827544e-01, 2.71349728e-01, 2.10650504e-01,\n "
" 4.24148560e-01, 9.78279829e-01, 5.26811719e-01,\n "
" 9.24481928e-01],\n "
" [ 7.44714022e-01, 2.27589488e-01, 7.76720643e-02,\n "
" 9.93781269e-01, 4.39466655e-01, 6.36421442e-01,\n "
" 8.98520589e-01],\n "
" [ 1.17154926e-01, 3.75053763e-01, 3.61411572e-02,\n "
" 1.01459026e-03, 7.44913578e-01, 2.32848793e-01,\n "
" 5.45857191e-01],\n "
" [ 1.76548839e-01, 3.47660542e-01, 3.47778410e-01,\n "
" 1.84943974e-01, 3.87695134e-02, 9.55624998e-01,\n "
" 4.20724988e-01],\n "
" [ 6.57340050e-01, 7.64973700e-01, 9.89420772e-01,\n "
" 3.95631552e-01, 4.33528155e-01, 7.69716501e-02,\n "
" 7.47923851e-01]], dtype=float32)"
]
},
"execution_count" : 276 ,
"metadata" : {},
"output_type" : " execute_result"
}
],
"source" : [
" h3[0][0][0][0] = -2.0\n "
" h3[0][0]"
]
},
{
"cell_type" : " code"
"execution_count" : 264 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [
{
"data" : {
"text/plain" : [
" array([[ 0.66239929, 0.13344583, 0.91010243, 0.1460737 , 0.40128991,\n "
" 0.51702583, 0.57517201],\n "
" [ 0.04876372, 0.29020485, 0.99930859, 0.93551981, 0.05357945,\n "
" 0.86496329, 0.99785376],\n "
" [ 0.82082754, 0.27134973, 0.2106505 , 0.42414856, 0.97827983,\n "
" 0.52681172, 0.92448193],\n "
" [ 0.74471402, 0.22758949, 0.07767206, 0.99378127, 0.43946666,\n "
" 0.63642144, 0.89852059],\n "
" [ 0.11715493, 0.37505376, 0.03614116, 0.00101459, 0.74491358,\n "
" 0.23284879, 0.54585719],\n "
" [ 0.17654884, 0.34766054, 0.34777841, 0.18494397, 0.03876951,\n "
" 0.955625 , 0.42072499],\n "
" [ 0.65734005, 0.7649737 , 0.98942077, 0.39563155, 0.43352816,\n "
" 0.07697165, 0.74792385]], dtype=float32)"
]
},
"execution_count" : 264 ,
"metadata" : {},
"output_type" : " execute_result"
}
],
"source" : [
" h2.data[0][0]"
]
},
{
"cell_type" : " code"
"execution_count" : 246 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [
{
"data" : {
"text/plain" : [
" (12740, (10, 26, 7, 7))"
]
},
"execution_count" : 246 ,
"metadata" : {},
"output_type" : " execute_result"
}
],
"source" : [
" (_pprob == _pprob2).sum(), _pprob.shape"
]
},
{
"cell_type" : " code"
"execution_count" : 97 ,
@@ -1397,25 +1604,67 @@
},
{
"cell_type" : " code"
"execution_count" : 70 ,
"execution_count" : 214 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [
{
"data" : {
"text/plain" : [
" (26, 7, 7)"
" array([ 0.43308464, 0.11953532, 0.29049623, 0.44944125, 0.25558978,\n "
" 0.3585023 , 0.61695206, 0.02689682, 0.11112458, 0.61580402,\n "
" 0.56997931, 0.36351889, 0.38775748, 0.39617181, 0.33856669,\n "
" 0.36801514, 0.23804198, 0.40024301, 0.09726732, 0.30708119,\n "
" 0.60552824, 0.6390807 , 0.49021432, 0.46792892, 0.08237838,\n "
" 0.62170118], dtype=float32)"
]
},
"execution_count" : 70 ,
"execution_count" : 214 ,
"metadata" : {},
"output_type" : " execute_result"
}
],
"source" : [
" _pobj = (_pprob[0] * _pconf[0]).transpose(1,2,0)\n "
" _pobj[1,3]"
]
},
{
"cell_type" : " code"
"execution_count" : 286 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [
{
"data" : {
"text/plain" : [
" array([[ 0.5436852 , 0.87299973, 0.54632992, 0.85493296, 0.79055774,\n "
" 0.04883667, 0.82923347],\n "
" [ 0.97029114, 0.99650782, 0.85720515, 0.29600522, 0.2879642 ,\n "
" 0.67986673, 0.996961 ],\n "
" [ 0.26799804, 0.19504192, 0.28671658, 0.86294001, 0.85752749,\n "
" 0.93175137, 0.04307476],\n "
" [ 0.01909542, 0.21990483, 0.97203565, 0.97640979, 0.07439857,\n "
" 0.90765405, 0.97738111],\n "
" [ 0.04907788, 0.57511681, 0.99103165, 0.30767789, 0.16018122,\n "
" 0.14995499, 0.72690445],\n "
" [ 0.89716494, 0.43719345, 0.94334656, 0.80119491, 0.32274631,\n "
" 0.61169893, 0.61864322],\n "
" [ 0.45193967, 0.77233005, 0.51528579, 0.10030524, 0.13502592,\n "
" 0.06036985, 0.98146945]], dtype=float32)"
]
},
"execution_count" : 286 ,
"metadata" : {},
"output_type" : " execute_result"
}
],
"source" : [
" _class_probs = _pprob[0] * _pconf[0]\n "
" _class_probs.shape"
" _max_class_probs_map = _class_probs.max(axis=0)\n "
" _max_class_probs_map"
]
},
{
@@ -2036,6 +2285,260 @@
" for x in np.arange(0,5)]"
]
},
{
"cell_type" : " code"
"execution_count" : 178 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [
{
"ename" : " TypeError"
"evalue" : " an integer is required"
"output_type" : " error"
"traceback" : [
" \u001b [0;31m---------------------------------------------------------------------------\u001b [0m"
" \u001b [0;31mTypeError\u001b [0m Traceback (most recent call last)"
" \u001b [0;32m<ipython-input-178-35a1f25e62ce>\u001b [0m in \u001b [0;36m<module>\u001b [0;34m()\u001b [0m\n \u001b [0;32m----> 1\u001b [0;31m \u001b [0mnp\u001b [0m\u001b [0;34m.\u001b [0m\u001b [0mrandom\u001b [0m\u001b [0;34m.\u001b [0m\u001b [0mrand\u001b [0m\u001b [0;34m(\u001b [0m\u001b [0;34m(\u001b [0m\u001b [0;36m1\u001b [0m\u001b [0;34m,\u001b [0m\u001b [0;36m3\u001b [0m\u001b [0;34m,\u001b [0m\u001b [0;36m5\u001b [0m\u001b [0;34m,\u001b [0m\u001b [0;36m5\u001b [0m\u001b [0;34m)\u001b [0m\u001b [0;34m)\u001b [0m\u001b [0;34m.\u001b [0m\u001b [0mshape\u001b [0m\u001b [0;34m\u001b [0m\u001b [0m\n \u001b [0m"
" \u001b [0;32mmtrand.pyx\u001b [0m in \u001b [0;36mmtrand.RandomState.rand (numpy/random/mtrand/mtrand.c:18928)\u001b [0;34m()\u001b [0m\n "
" \u001b [0;32mmtrand.pyx\u001b [0m in \u001b [0;36mmtrand.RandomState.random_sample (numpy/random/mtrand/mtrand.c:14895)\u001b [0;34m()\u001b [0m\n "
" \u001b [0;32mmtrand.pyx\u001b [0m in \u001b [0;36mmtrand.cont0_array (numpy/random/mtrand/mtrand.c:2383)\u001b [0;34m()\u001b [0m\n "
" \u001b [0;31mTypeError\u001b [0m: an integer is required"
]
}
],
"source" : [
" np.random.rand((1,3,5,5)).shape"
]
},
{
"cell_type" : " code"
"execution_count" : 306 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [
{
"data" : {
"text/plain" : [
" array([2, 3, 4, 5])"
]
},
"execution_count" : 306 ,
"metadata" : {},
"output_type" : " execute_result"
}
],
"source" : [
" a = np.arange(1, 6)\n "
" a[1:]"
]
},
{
"cell_type" : " code"
"execution_count" : 217 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [
{
"name" : " stdout"
"output_type" : " stream"
"text" : [
" 1 2\n "
" 2 3\n "
" 3 4\n "
" 4 5\n "
]
}
],
"source" : [
" for i, item in enumerate(a[1:], 1):\n "
" print(i, item)"
]
},
{
"cell_type" : " code"
"execution_count" : 299 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [
{
"data" : {
"text/plain" : [
" array([11, 12, 13, 14, 15])"
]
},
"execution_count" : 299 ,
"metadata" : {},
"output_type" : " execute_result"
}
],
"source" : [
" b = np.arange(11, 16)\n "
" b"
]
},
{
"cell_type" : " code"
"execution_count" : 296 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [
{
"name" : " stdout"
"output_type" : " stream"
"text" : [
" 1 1 11\n "
" 2 2 12\n "
" 3 3 13\n "
" 4 4 14\n "
" 5 5 15\n "
]
}
],
"source" : [
" for i, (a, b) in enumerate(zip(a, b), 1):\n "
" print (i, a, b)"
]
},
{
"cell_type" : " code"
"execution_count" : 304 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [
{
"data" : {
"text/plain" : [
" array([15, 14, 13, 12, 11])"
]
},
"execution_count" : 304 ,
"metadata" : {},
"output_type" : " execute_result"
}
],
"source" : [
" np.sort(b)[::-1]"
]
},
{
"cell_type" : " code"
"execution_count" : 309 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [
{
"name" : " stdout"
"output_type" : " stream"
"text" : [
" 1 11\n "
" 1 12\n "
" 1 13\n "
" 1 14\n "
" 1 15\n "
" 2 11\n "
" 2 12\n "
" 2 13\n "
" 2 14\n "
" 2 15\n "
" 3 11\n "
" 3 12\n "
" 3 13\n "
" 3 14\n "
" 3 15\n "
" 4 11\n "
" 4 12\n "
" 4 13\n "
" 4 14\n "
" 4 15\n "
" 5 11\n "
" 5 12\n "
" 5 13\n "
" 5 14\n "
" 5 15\n "
]
}
],
"source" : [
" import itertools\n "
" for _a, _b in itertools.product(a, b):\n "
" print(_a, _b)"
]
},
{
"cell_type" : " code"
"execution_count" : 313 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [
{
"data" : {
"text/plain" : [
" list"
]
},
"execution_count" : 313 ,
"metadata" : {},
"output_type" : " execute_result"
}
],
"source" : [
" a = [1,2,3,4,5]\n "
" type(a)"
]
},
{
"cell_type" : " code"
"execution_count" : 320 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [
{
"data" : {
"text/plain" : [
" 0.5"
]
},
"execution_count" : 320 ,
"metadata" : {},
"output_type" : " execute_result"
}
],
"source" : [
" a = [0.2, 0.1, 0.5, 0.9, 0.8]\n "
" np.asarray(a).mean()"
]
},
{
"cell_type" : " code"
"execution_count" : 321 ,
"metadata" : {
"collapsed" : false
},
"outputs" : [
{
"data" : {
"text/plain" : [
" 5250.0"
]
},
"execution_count" : 321 ,
"metadata" : {},
"output_type" : " execute_result"
}
],
"source" : [
" 750 * (3.5+3.5)"
]
},
{
"cell_type" : " code"
"execution_count" : null ,