Article <mailman.49.1722372595.2981.python-list@python.org>

Deutsch English Français Italiano
<mailman.49.1722372595.2981.python-list@python.org>

View for Bookmarking (what is this?)
Look up another Usenet article
Path: ...!news.mixmin.net!news.swapon.de!fu-berlin.de!uni-berlin.de!not-for-mail
From: marc nicole <mk1853387@gmail.com>
Newsgroups: comp.lang.python
Subject: Re: Predicting an object over an pretrained model is not working
Date: Tue, 30 Jul 2024 22:49:21 +0200
Lines: 269
Message-ID: <mailman.49.1722372595.2981.python-list@python.org>
References: <CAGJtH9Qjv2fQm=_HKwhoGS11uh+u4YoTVzYGHF=2jZC9HpdV9A@mail.gmail.com>
 <263356ef-7ad8-4abc-9940-bd8536ee13eb@tompassin.net>
 <CAGJtH9TnoFa_JJNi=E0oDouKZjq_sfGYmr0WOFOfZtaGcQTyXA@mail.gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Cancel-Lock: sha1:0+luE/LnbzDo9LGWEmO9KljyGNE= sha256:8rwBoI9ZfcYKPq3IGICKRJSfzFB7va5EaRxvJOSV/bU=
In-Reply-To: <263356ef-7ad8-4abc-9940-bd8536ee13eb@tompassin.net>
Precedence: list
Bytes: 17802

OK, but how's the probability of small_ball greater than others? I can't
find it anyway, what's its value?

Le mar. 30 juil. 2024 =C3=A0 21:37, Thomas Passin via Python-list <
python-list@python.org> a =C3=A9crit :

> On 7/30/2024 2:18 PM, marc nicole via Python-list wrote:
> > Hello all,
> >
> > I want to predict an object by given as input an image and want to have
> my
> > model be able to predict the label. I have trained a model using
> tensorflow
> > based on annotated database where the target object to predict was adde=
d
> to
> > the pretrained model. the code I am using is the following where I set
> the
> > target object image as input and want to have the prediction output:
> >
> >
> >
> >
> >
> >
> >
> >
> > class MultiObjectDetection():
> >
> >      def __init__(self, classes_name):
> >
> >          self._classes_name =3D classes_name
> >          self._num_classes =3D len(classes_name)
> >
> >          self._common_params =3D {'image_size': 448, 'num_classes':
> > self._num_classes,
> >                  'batch_size':1}
> >          self._net_params =3D {'cell_size': 7, 'boxes_per_cell':2,
> > 'weight_decay': 0.0005}
> >          self._net =3D YoloTinyNet(self._common_params, self._net_param=
s,
> > test=3DTrue)
> >
> >      def predict_object(self, image):
> >          predicts =3D self._net.inference(image)
> >          return predicts
> >
> >      def process_predicts(self, resized_img, predicts, thresh=3D0.2):
> >          """
> >          process the predicts of object detection with one image input.
> >
> >          Args:
> >              resized_img: resized source image.
> >              predicts: output of the model.
> >              thresh: thresh of bounding box confidence.
> >          Return:
> >              predicts_dict: {"stick": [[x1, y1, x2, y2, scores1],
> [...]]}.
> >          """
> >          cls_num =3D self._num_classes
> >          bbx_per_cell =3D self._net_params["boxes_per_cell"]
> >          cell_size =3D self._net_params["cell_size"]
> >          img_size =3D self._common_params["image_size"]
> >          p_classes =3D predicts[0, :, :, 0:cls_num]
> >          C =3D predicts[0, :, :, cls_num:cls_num+bbx_per_cell] # two
> > bounding boxes in one cell.
> >          coordinate =3D predicts[0, :, :, cls_num+bbx_per_cell:] # all
> > bounding boxes position.
> >
> >          p_classes =3D np.reshape(p_classes, (cell_size, cell_size, 1,
> cls_num))
> >          C =3D np.reshape(C, (cell_size, cell_size, bbx_per_cell, 1))
> >
> >          P =3D C * p_classes # confidencefor all classes of all boundin=
g
> > boxes (cell_size, cell_size, bounding_box_num, class_num) =3D (7, 7, 2,
> > 1).
> >
> >          predicts_dict =3D {}
> >          for i in range(cell_size):
> >              for j in range(cell_size):
> >                  temp_data =3D np.zeros_like(P, np.float32)
> >                  temp_data[i, j, :, :] =3D P[i, j, :, :]
> >                  position =3D np.argmax(temp_data) # refer to the class
> > num (with maximum confidence) for every bounding box.
> >                  index =3D np.unravel_index(position, P.shape)
> >
> >                  if P[index] > thresh:
> >                      class_num =3D index[-1]
> >                      coordinate =3D np.reshape(coordinate, (cell_size,
> > cell_size, bbx_per_cell, 4)) # (cell_size, cell_size,
========== REMAINDER OF ARTICLE TRUNCATED ==========