Q&A
質問させていただきます．
現在私は，tensorflow の"object detection API" を利用しています．
利用しているのはgit上にある以下のプログラムです．
そこで私は 物体認識をして画像上に表示される bouding box の座標をしりたいと思っています．
通常 opencv などで行った際には (boundingboxの左上のx座標,boundingboxの左上のy座標, 横の長さ,縦の長さ)のように値が保持してあります．
プログラミングコード上でおそらく関係があると思われる Detection の部分にある"boxes"などの中身を調べてみました．
以下がその近辺のコードです．
python
1with detection_graph.as_default(): 2 with tf.Session(graph=detection_graph) as sess: 3 # Definite input and output Tensors for detection_graph 4 image_tensor = detection_graph.get_tensor_by_name('image_tensor:0') 5 # Each box represents a part of the image where a particular object was detected. 6 detection_boxes = detection_graph.get_tensor_by_name('detection_boxes:0') 7 # Each score represent how level of confidence for each of the objects. 8 # Score is shown on the result image, together with the class label. 9 detection_scores = detection_graph.get_tensor_by_name('detection_scores:0') 10 detection_classes = detection_graph.get_tensor_by_name('detection_classes:0') 11 num_detections = detection_graph.get_tensor_by_name('num_detections:0') 12 for image_path in TEST_IMAGE_PATHS: 13 image = Image.open(image_path) 14 # the array based representation of the image will be used later in order to prepare the 15 # result image with boxes and labels on it. 16 image_np = load_image_into_numpy_array(image) 17 # Expand dimensions since the model expects images to have shape: [1, None, None, 3] 18 image_np_expanded = np.expand_dims(image_np, axis=0) 19 # Actual detection. 20 (boxes, scores, classes, num) = sess.run( 21 [detection_boxes, detection_scores, detection_classes, num_detections], 22 feed_dict={image_tensor: image_np_expanded}) 23 # Visualization of the results of a detection. 24 vis_util.visualize_boxes_and_labels_on_image_array( 25 image_np, 26 np.squeeze(boxes), 27 np.squeeze(classes).astype(np.int32), 28 np.squeeze(scores), 29 category_index, 30 use_normalized_coordinates=True, 31 line_thickness=8) 32 plt.figure(figsize=IMAGE_SIZE) 33 plt.imshow(image_np)
以下が結果になります．
しかしながら，直接boudingbox の座標に関係があるのかが分かりませんでした．
他の変数の中身も見てみましたが，座標に関係のあるものが無いように思えました．
result
1[[[ 3.70766222e-02 2.32391059e-02 8.62020969e-01 3.18441600e-01] 2 [ 1.00141376e-01 3.91916126e-01 9.58822250e-01 9.73572135e-01] 3 [ 5.80091894e-01 4.02289629e-01 9.10106719e-01 6.81827903e-01] 4 [ 2.67864764e-03 5.37680089e-01 2.50190973e-01 6.97518170e-01] 5 [ 6.59326911e-02 3.71420026e-01 9.47743475e-01 9.82712150e-01] 6 [ 2.68082619e-02 7.62826264e-01 6.32788718e-01 9.91576493e-01] 7 [ 1.74010396e-02 3.69260252e-01 8.60050261e-01 6.47671759e-01] 8 [ 2.62468457e-02 2.15394646e-01 7.13022113e-01 6.44605398e-01] 9 [ 1.07638836e-01 5.58825672e-01 7.33044744e-01 7.30897963e-01] 10 [ 3.79294306e-02 1.11420602e-01 3.52208912e-01 3.05902272e-01] 11 [ 0.00000000e+00 4.11696434e-01 6.22777462e-01 6.51909351e-01] 12 [ 2.79582739e-02 6.83753192e-03 4.81684983e-01 3.11877012e-01] 13 [ 7.44389296e-02 2.71154195e-01 9.00867224e-01 9.64441538e-01] 14 [ 4.99033928e-03 2.87592858e-01 6.27175033e-01 5.26544154e-01] 15 [ 4.67826426e-03 3.77882779e-01 3.10840487e-01 6.87657654e-01] 16 [ 1.08135045e-02 2.45042488e-01 8.66623044e-01 4.76782322e-01] 17 [ 5.08013368e-02 0.00000000e+00 8.92649353e-01 5.29527903e-01] 18 [ 6.39598444e-03 6.85519636e-01 8.14009309e-02 7.63971627e-01] 19 [ 7.34354928e-03 7.00937867e-01 8.37567449e-02 7.99702287e-01] 20 [ 9.56903398e-03 7.89382577e-01 2.77421892e-01 9.53748822e-01] 21 [ 1.14383996e-02 2.46351182e-01 7.93706298e-01 4.77306306e-01] 22 [ 1.47813708e-02 2.62614429e-01 2.61748374e-01 7.37113655e-01] 23 [ 2.08409131e-03 5.27146995e-01 2.99976110e-01 7.69435227e-01] 24 [ 6.45412028e-01 5.97376883e-01 7.72994697e-01 7.08751142e-01] 25 [ 1.76023692e-01 8.01044524e-01 2.56199092e-01 8.36520731e-01] 26 [ 1.77701205e-01 8.30787718e-01 2.77940273e-01 8.98237050e-01] 27 [ 4.49800789e-01 4.20949638e-01 8.72183025e-01 6.59005344e-01] 28 [ 4.21552211e-02 7.85431147e-01 3.13383758e-01 9.55286980e-01] 29 [ 0.00000000e+00 5.06608784e-02 7.51202464e-01 5.64320087e-01] 30 [ 2.52031386e-02 1.14164032e-01 6.10430717e-01 3.33315998e-01] 31 [ 8.68688524e-03 9.93093550e-02 2.13545889e-01 3.07423383e-01] 32 [ 9.36175287e-02 7.85452187e-01 4.23507959e-01 9.72094476e-01] 33 [ 0.00000000e+00 5.60392439e-01 6.07045770e-01 7.04684317e-01] 34 [ 1.60332471e-02 3.91475856e-04 4.99358475e-01 1.80552453e-01] 35 [ 6.36043251e-02 3.12048316e-01 8.48924041e-01 5.44608653e-01] 36 [ 5.27268648e-03 6.59941792e-01 2.47535557e-01 7.97456503e-01] 37 [ 1.02197461e-01 7.63699412e-01 3.39428455e-01 9.21994090e-01] 38 [ 6.34029388e-01 4.12018538e-01 7.77234197e-01 7.11640954e-01] 39 [ 4.99218106e-02 5.12819707e-01 6.33476317e-01 7.85471737e-01] 40 [ 1.41868293e-02 9.28779602e-01 4.37340528e-01 9.98276949e-01] 41 [ 8.84260982e-04 5.37971973e-01 8.48287344e-02 7.90319443e-01] 42 [ 1.74092352e-02 3.24149430e-01 5.37369490e-01 6.78354681e-01] 43 [ 8.63035023e-03 6.22962892e-01 3.05343926e-01 8.98170769e-01] 44 [ 5.36475003e-01 3.25907648e-01 9.20678198e-01 7.30400264e-01] 45 [ 1.32080913e-03 7.76022434e-01 3.40186447e-01 9.96103764e-01] 46 [ 0.00000000e+00 6.62098348e-01 8.21765587e-02 9.03941929e-01] 47 [ 1.66808799e-01 7.63557017e-01 4.42327499e-01 9.08092916e-01] 48 [ 1.39167011e-02 9.07135189e-01 3.12922806e-01 1.00000000e+00] 49 [ 1.57200694e-02 8.35942447e-01 7.62703001e-01 9.97816980e-01] 50 [ 9.23319161e-03 4.49134111e-01 2.72807360e-01 8.67401481e-01] 51 [ 6.45412028e-01 5.97376883e-01 7.72994697e-01 7.08751142e-01] 52 [ 0.00000000e+00 6.07310951e-01 8.00046027e-02 8.40086758e-01] 53 [ 6.11052752e-01 6.06802702e-01 7.49152660e-01 7.06053734e-01] 54 [ 2.38389701e-01 0.00000000e+00 7.98261642e-01 7.60288477e-01] 55 [ 3.70766222e-02 2.32391059e-02 8.62020969e-01 3.18441600e-01] 56 [ 8.88931751e-03 0.00000000e+00 3.26228440e-01 3.33230644e-01] 57 [ 2.04673111e-02 4.26852524e-01 4.37012225e-01 6.71041191e-01] 58 [ 0.00000000e+00 4.76602226e-01 2.65630186e-01 6.75271511e-01] 59 [ 1.80710256e-02 2.36390218e-01 3.94988418e-01 7.23949134e-01] 60 [ 0.00000000e+00 3.33192140e-01 2.98385859e-01 5.26110590e-01] 61 [ 1.00441381e-01 6.81784824e-02 2.73970127e-01 3.16320151e-01] 62 [ 7.04130530e-03 0.00000000e+00 6.89402938e-01 4.04221177e-01] 63 [ 2.88287550e-01 0.00000000e+00 6.87736273e-01 5.00128388e-01] 64 [ 1.72952354e-01 7.29220152e-01 2.93541282e-01 9.44286823e-01] 65 [ 1.17731392e-02 3.24624926e-01 2.20594078e-01 5.77684522e-01] 66 [ 6.86677277e-01 4.09583986e-01 8.14984739e-01 7.10663617e-01] 67 [ 2.02762514e-01 6.69540823e-01 3.74175936e-01 9.30079639e-01] 68 [ 1.18174672e-01 0.00000000e+00 3.58328700e-01 3.40447336e-01] 69 [ 1.63917422e-01 8.00174356e-01 5.13882160e-01 9.61729050e-01] 70 [ 2.43298978e-01 2.45772272e-01 7.61221290e-01 1.00000000e+00] 71 [ 0.00000000e+00 3.50008696e-01 6.97703511e-02 6.64999008e-01] 72 [ 0.00000000e+00 6.56346321e-01 1.34895504e-01 7.82198906e-01] 73 [ 6.88554347e-03 2.51892745e-01 3.99504662e-01 4.55886722e-01] 74 [ 0.00000000e+00 6.88617468e-01 1.44334793e-01 8.23197842e-01] 75 [ 0.00000000e+00 3.20033640e-01 1.47724807e-01 5.30816138e-01] 76 [ 0.00000000e+00 6.29961789e-02 3.88361305e-01 3.91400486e-01] 77 [ 0.00000000e+00 8.28370631e-01 2.65981734e-01 9.88341630e-01] 78 [ 1.82754733e-03 3.41375083e-01 5.76723889e-02 4.78018731e-01] 79 [ 0.00000000e+00 3.58040571e-01 1.53169647e-01 6.13790154e-01] 80 [ 5.31935155e-01 3.84203613e-01 7.78207958e-01 7.21878469e-01] 81 [ 6.36111915e-01 3.22093010e-01 7.82574832e-01 5.45183182e-01] 82 [ 58412367e-01 6.44178033e-01 7.31922507e-01 8.87382388e-01] 83 [ 5.28672636e-02 2.11526453e-03 7.16688514e-01 1.32305130e-01] 84 [ 1.46860301e-01 8.43356252e-01 5.86594760e-01 9.81495738e-01] 85 [ 6.22597337e-03 1.72302663e-01 5.41191220e-01 8.17064404e-01] 86 [ 9.56874192e-02 0.00000000e+00 7.61249065e-01 9.89065170e-01] 87 [ 6.58449531e-03 9.32844043e-01 1.69511288e-01 9.95045900e-01] 88 [ 1.99106708e-03 6.38102531e-01 6.89450651e-02 7.14508414e-01] 89 [ 0.00000000e+00 7.84308791e-01 7.88945258e-02 1.00000000e+00] 90 [ 1.57109201e-02 3.17976117e-01 4.13692623e-01 4.79996681e-01] 91 [ 7.31878519e-01 1.56241953e-02 9.86946940e-01 9.25869226e-01] 92 [ 2.26111412e-02 2.12461650e-01 9.71305549e-01 8.51539791e-01] 93 [ 4.04597819e-02 5.12805581e-02 8.00263524e-01 9.93615806e-01] 94 [ 3.79322648e-01 5.40679693e-01 8.81464958e-01 9.43988562e-01] 95 [ 1.76028907e-02 3.19518447e-01 4.30800200e-01 4.78415072e-01] 96 [ 0.00000000e+00 1.83706462e-01 2.29834288e-01 5.22784472e-01] 97 [ 1.68697685e-02 2.61039078e-01 4.68145788e-01 5.87137640e-01] 98 [ 6.86408281e-01 3.27755451e-01 8.19596648e-01 5.54131150e-01] 99 [ 2.23855555e-01 4.12261397e-01 6.54340148e-01 7.17972040e-01] 100 [ 0.00000000e+00 8.49669576e-01 1.65976733e-01 9.89492893e-01]]] 101
ちなみに物体認識をした画像についてはsample として使用されている以下の画像を利用しました．
bouding box の座標の調べ方がわかる方がおりましたらよろしくお願いします．
回答1件
0
ベストアンサー
元のコードを見てみるとよいです。
tensorflow/models/blob/master/research/object_detection/utils/visualization_utils.py#L385
Python
1# Draw all boxes onto image. 2for box, color in box_to_color_map.items(): 3 ymin, xmin, ymax, xmax = box
こうあるので、boxの中身の要素が何を意味するのかわかります。
ついでなので、実際に描画する部分も見てみましょう。
tensorflow/models/blob/master/research/object_detection/utils/visualization_utils.py#L124
Python
1(left, right, top, bottom) = (xmin * im_width, xmax * im_width, 2 ymin * im_height, ymax * im_height)
これを真似ればいいですね。
投稿2017/11/08 10:54
編集2017/11/08 10:56
総合スコア35660
あなたの回答
tips
プレビュー
バッドをするには、ログインかつ
こちらの条件を満たす必要があります。
2017/11/08 11:49