diff --git a/.github/workflows/CI-responsibleai-text-vision-pytest.yml b/.github/workflows/CI-responsibleai-text-vision-pytest.yml
index b762c56083..0c420e69ca 100644
--- a/.github/workflows/CI-responsibleai-text-vision-pytest.yml
+++ b/.github/workflows/CI-responsibleai-text-vision-pytest.yml
@@ -88,7 +88,9 @@ jobs:
       - if: ${{ (matrix.packageDirectory == 'responsibleai_vision') && ((matrix.pythonVersion == '3.7') || (matrix.pythonVersion == '3.8')) }}
         name: Install automl dependencies
         shell: bash -l {0}
+        # pycocotools is a downstream dependency of automl requirements, but fails to install with pip
         run: |
+          conda install pycocotools==2.0.4 -c conda-forge
           pip install -r ${{ matrix.packageDirectory }}/requirements-automl.txt
 
       - name: Install package
diff --git a/responsibleai_vision/requirements.txt b/responsibleai_vision/requirements.txt
index ab5b5077dc..97e350b85a 100644
--- a/responsibleai_vision/requirements.txt
+++ b/responsibleai_vision/requirements.txt
@@ -1,5 +1,7 @@
 numpy>=1.17.2
 pandas>=0.25.1,<2.0.0 # TODO: remove ceiling on version.
+Pillow>=10.0.0; python_version>"3.7" # due to breaking changes in v10.0.0 (https://pillow.readthedocs.io/en/latest/releasenotes/10.0.0.html)
+Pillow<10.0.0; python_version<="3.7" # Pillow v10.0.0 is only available starting with Python 3.8
 scikit-learn>=0.22.1
 scipy>=1.4.1
 semver~=2.13.0
diff --git a/responsibleai_vision/responsibleai_vision/managers/explainer_manager.py b/responsibleai_vision/responsibleai_vision/managers/explainer_manager.py
index 83c627b895..6b95eca14d 100644
--- a/responsibleai_vision/responsibleai_vision/managers/explainer_manager.py
+++ b/responsibleai_vision/responsibleai_vision/managers/explainer_manager.py
@@ -266,14 +266,23 @@ def compute_single_explanation(self,
                 # get_drise_saliency_map only recognizes GPU and CPU
                 if device == Device.AUTO.value:
                     device = None
-                fl, _, _, = get_drise_saliency_map(img,
-                                                   self._model,
-                                                   len(self._classes),
-                                                   savename=str(index),
-                                                   nummasks=self._num_masks,
-                                                   maskres=mask_res_tuple,
-                                                   devicechoice=device,
-                                                   max_figures=5000)
+                try:
+                    fl, _, _, = get_drise_saliency_map(
+                        img,
+                        self._model,
+                        len(self._classes),
+                        savename=str(index),
+                        nummasks=self._num_masks,
+                        maskres=mask_res_tuple,
+                        devicechoice=device,
+                        max_figures=5000)
+                except ValueError as ve:
+                    if str(ve) == 'No detections found':
+                        raise UserConfigValidationException(
+                            "No detections found in image. "
+                            "Please increase num_masks or mask_res.")
+                    else:
+                        raise ve
                 if object_index is None:
                     return fl
                 b64_string = fl[object_index]
@@ -577,11 +586,13 @@ def _is_object_detection_task(self):
         return self._task_type == ModelTask.OBJECT_DETECTION
 
     def _get_fail_str(self):
-        fail = Image.new('RGB', (100, 100))
+        fail = Image.new('RGB', (250, 250))
         draw = ImageDraw.Draw(fail)
         font = ImageFont.load_default()
         text = "saliency map could not be created"
-        textwidth, textheight = draw.textsize(text, font)
+        left, top, right, bottom = draw.textbbox((0, 0), text, font)
+        textwidth = right - left
+        textheight = bottom - top
         x = (fail.width - textwidth) // 2
         y = (fail.height - textheight) // 2
         draw.text((x, y), text, fill="white", font=font)