示例#1
0
    def __next__(self):
        """
        Returns:
            draw_blend_generator output, deblender output and measurement output.
        """

        blend_output = next(self.draw_blend_generator)
        deblend_results = {}
        measured_results = {}
        input_args = [(blend_output, i) for i in range(self.batch_size)]
        batch_results = multiprocess(
            self.run_batch,
            input_args,
            self.cpus,
            self.multiprocessing,
            self.verbose,
        )
        for i in range(self.batch_size):
            deblend_results.update({i: batch_results[i][0]})
            measured_results.update({i: batch_results[i][1]})
        if self.verbose:
            print("Measurement performed on batch")
        return blend_output, deblend_results, measured_results
示例#2
0
    def __next__(self):
        """Outputs dictionary containing blend output (images and catalogs) in batches.

        Returns:
            output: Dictionary with blend images, isolated object images, blend catalog,
            PSF images and WCS.
        """
        blend_list = {}
        blend_images = {}
        isolated_images = {}
        blend_cat = next(self.blend_generator)
        mini_batch_size = np.max([self.batch_size // self.cpus, 1])
        psfs = {}
        wcss = {}

        for s in self.surveys:
            pix_stamp_size = int(self.stamp_size / s.pixel_scale)

            # make PSF and WCS
            psf = []
            for filt in s.filters:
                if callable(filt.psf):
                    generated_psf = filt.psf(
                    )  # generate the PSF with the provided function
                    if isinstance(generated_psf, galsim.GSObject):
                        psf.append(generated_psf)
                    else:
                        raise TypeError(
                            f"The generated PSF with the provided function"
                            f"for filter '{filt.name}' is not a galsim object")
                elif isinstance(filt.psf, galsim.GSObject):
                    psf.append(filt.psf)  # or directly retrieve the PSF
                else:
                    raise TypeError(
                        f"The PSF within filter '{filt.name}' is neither a "
                        f"function nor a galsim object")
            wcs = make_wcs(s.pixel_scale, (pix_stamp_size, pix_stamp_size))
            psfs[s.name] = psf
            wcss[s.name] = wcs

            input_args = []
            seedseq_minibatch = self.seedseq.spawn(self.batch_size //
                                                   mini_batch_size + 1)

            for i in range(0, self.batch_size, mini_batch_size):
                cat = copy.deepcopy(blend_cat[i:i + mini_batch_size])
                input_args.append((cat, psf, wcs, s,
                                   seedseq_minibatch[i // mini_batch_size]))

            # multiprocess and join results
            # ideally, each cpu processes a single mini_batch
            mini_batch_results = multiprocess(
                self.render_mini_batch,
                input_args,
                cpus=self.cpus,
                verbose=self.verbose,
            )

            # join results across mini-batches.
            batch_results = list(chain(*mini_batch_results))

            # decide image_shape based on channels_last bool.
            option1 = (len(s.filters), pix_stamp_size, pix_stamp_size)
            option2 = (pix_stamp_size, pix_stamp_size, len(s.filters))
            image_shape = option1 if not self.channels_last else option2

            # organize results.
            blend_images[s.name] = np.zeros((self.batch_size, *image_shape))
            isolated_images[s.name] = np.zeros(
                (self.batch_size, self.max_number, *image_shape))
            blend_list[s.name] = []
            for i in range(self.batch_size):
                blend_images[s.name][i] = batch_results[i][0]
                isolated_images[s.name][i] = batch_results[i][1]
                blend_list[s.name].append(batch_results[i][2])

            # save results if requested.
            if self.save_path is not None:
                if not os.path.exists(os.path.join(self.save_path, s.name)):
                    os.mkdir(os.path.join(self.save_path, s.name))

                np.save(os.path.join(self.save_path, s.name, "blended"),
                        blend_images[s.name])
                np.save(os.path.join(self.save_path, s.name, "isolated"),
                        isolated_images[s.name])
                for i in range(len(batch_results)):
                    blend_list[s.name][i].write(
                        os.path.join(self.save_path, s.name,
                                     f"blend_info_{i}"),
                        format="ascii",
                        overwrite=True,
                    )
        if self.is_multiresolution:
            output = {
                "blend_images": blend_images,
                "isolated_images": isolated_images,
                "blend_list": blend_list,
                "psf": psfs,
                "wcs": wcss,
            }
        else:
            survey_name = self.surveys[0].name
            output = {
                "blend_images": blend_images[survey_name],
                "isolated_images": isolated_images[survey_name],
                "blend_list": blend_list[survey_name],
                "psf": psfs[survey_name],
                "wcs": wcss[survey_name],
            }
        return output
    def __next__(self):
        """
        Returns:
            Dictionary with blend images, isolated object images, blend catalog,
            and observing conditions.
        """
        batch_blend_cat, batch_obs_cond = {}, {}
        blend_images = {}
        isolated_images = {}
        for s in self.surveys:
            pix_stamp_size = int(self.stamp_size / s.pixel_scale)
            batch_blend_cat[s.name], batch_obs_cond[s.name] = [], []
            blend_images[s.name] = np.zeros((self.batch_size, pix_stamp_size,
                                             pix_stamp_size, len(s.bands)))
            isolated_images[s.name] = np.zeros((
                self.batch_size,
                self.max_number,
                pix_stamp_size,
                pix_stamp_size,
                len(s.bands),
            ))

        in_batch_blend_cat = next(self.blend_generator)
        obs_conds = next(
            self.observing_generator)  # same for every blend in batch.
        mini_batch_size = np.max([self.batch_size // self.cpus, 1])
        for s in self.surveys:
            input_args = [(
                copy.deepcopy(in_batch_blend_cat[i:i + mini_batch_size]),
                copy.deepcopy(obs_conds[s.name]),
                s,
            ) for i in range(0, self.batch_size, mini_batch_size)]

            # multiprocess and join results
            # ideally, each cpu processes a single mini_batch
            mini_batch_results = multiprocess(
                self.render_mini_batch,
                input_args,
                self.cpus,
                self.multiprocessing,
                self.verbose,
            )

            # join results across mini-batches.
            batch_results = list(chain(*mini_batch_results))

            # organize results.
            for i in range(self.batch_size):
                blend_images[s.name][i] = batch_results[i][0]
                isolated_images[s.name][i] = batch_results[i][1]
                batch_blend_cat[s.name].append(batch_results[i][2])
        if len(self.surveys) > 1:
            output = {
                "blend_images": blend_images,
                "isolated_images": isolated_images,
                "blend_list": batch_blend_cat,
                "obs_condition": obs_conds,
            }
        else:
            survey_name = self.surveys[0].name
            output = {
                "blend_images": blend_images[survey_name],
                "isolated_images": isolated_images[survey_name],
                "blend_list": batch_blend_cat[survey_name],
                "obs_condition": obs_conds[survey_name],
            }
        return output
示例#4
0
    def __next__(self):
        """Return measurement results on a single batch from the draw_blend_generator.

        Returns:
            draw_blend_generator output from its `__next__` method.
            measurement_results (dict): Dictionary with keys being the name of each
                `measure_function` passed in. Each value is a dictionary containing keys
                `catalog`, `deblended_images`, and `segmentation` storing the values returned by
                the corresponding measure_function` for one batch.
        """
        blend_output = next(self.draw_blend_generator)
        catalog = {}
        segmentation = {}
        deblended_images = {}
        for f in self.measure_functions:
            for m in range(len(self.measure_kwargs)):
                key_name = f.__name__ + str(m) if len(
                    self.measure_kwargs) > 1 else f.__name__
                catalog[key_name] = []
                segmentation[key_name] = []
                deblended_images[key_name] = []
        for m, measure_kwargs in enumerate(self.measure_kwargs):
            args_iter = ((blend_output, i) for i in range(self.batch_size))
            kwargs_iter = repeat(measure_kwargs)
            measure_output = multiprocess(
                self.run_batch,
                args_iter,
                kwargs_iter=kwargs_iter,
                cpus=self.cpus,
                verbose=self.verbose,
            )

            if self.verbose:
                print(f"Measurement {m} performed on batch")

            for i, f in enumerate(self.measure_functions):
                key_name = f.__name__ + str(m) if len(
                    self.measure_kwargs) > 1 else f.__name__
                for j in range(len(measure_output)):
                    catalog[key_name].append(measure_output[j][i].get(
                        "catalog", None))
                    segmentation[key_name].append(measure_output[j][i].get(
                        "segmentation", None))
                    deblended_images[key_name].append(measure_output[j][i].get(
                        "deblended_images", None))
                # If multiresolution, we reverse the order between the survey name and
                # the index of the blend
                if self.is_multiresolution:
                    survey_keys = list(blend_output["blend_list"].keys())
                    # We duplicate the catalog for each survey to get the pixel coordinates
                    catalogs_temp = {}
                    for surv in survey_keys:
                        catalogs_temp[surv] = add_pixel_columns(
                            catalog[key_name], blend_output["wcs"][surv])
                    catalog[key_name] = catalogs_temp

                    segmentation[key_name] = reverse_list_dictionary(
                        segmentation[key_name], survey_keys)
                    deblended_images[key_name] = reverse_list_dictionary(
                        deblended_images[key_name], survey_keys)

                else:
                    catalog[key_name] = add_pixel_columns(
                        catalog[key_name], blend_output["wcs"])

                # save results if requested.
                if self.save_path is not None:

                    if not os.path.exists(
                            os.path.join(self.save_path, key_name)):
                        os.mkdir(os.path.join(self.save_path, key_name))

                    if segmentation[key_name] is not None:
                        np.save(
                            os.path.join(self.save_path, key_name,
                                         "segmentation"),
                            segmentation[key_name],
                        )
                    if deblended_images[key_name] is not None:
                        np.save(
                            os.path.join(self.save_path, key_name,
                                         "deblended_images"),
                            deblended_images[key_name],
                        )
                    for j, cat in enumerate(catalog[key_name]):
                        cat.write(
                            os.path.join(self.save_path, key_name,
                                         f"detection_catalog_{j}"),
                            format="ascii",
                            overwrite=True,
                        )
        measure_results = {
            "catalog": catalog,
            "segmentation": segmentation,
            "deblended_images": deblended_images,
        }
        return blend_output, measure_results