Bootcamp - Catherine Jin #113

cathjin · 2024-10-02T02:11:18Z

No description provided.

maxlou05

Reviewed

maxlou05 · 2024-10-02T18:36:11Z

modules/bootcamp/detect_landing_pad.py

-
-        # Loop over the boxes list and create a list of bounding boxes
-        bounding_boxes = []
+        for r in predictions:


You don't really need to do this, since we only predicted 1 thing (image), so it will return only 1 results object (for our 1 image). The reason it has a list is because sometimes you run predict on a list of images, then it will return a list of results objects. https://docs.ultralytics.com/modes/predict/#working-with-results

maxlou05 · 2024-10-02T18:36:56Z

modules/bootcamp/detect_landing_pad.py

+
+            # Plot the annotated image from the Result object
+            # Include the confidence value
+            image_annotated = prediction.plot()


Please also include/print the confidence value onto the image

maxlou05 · 2024-10-02T18:38:03Z

modules/bootcamp/detect_landing_pad.py

+            # and convert to a numpy array
+            boxes_cpu = boxes_xyxy.detach().cpu().numpy()
+            # Loop over the boxes list and create a list of bounding boxes
+            for i in enumerate(boxes_cpu):


If you are using enumerate, the typical usage is for i, box in enumerate(boxes_cpu). You can tuple unpack in the for statement itself.

maxlou05 · 2024-10-02T18:38:54Z

modules/bootcamp/detect_landing_pad.py

+            for i in enumerate(boxes_cpu):
+                index = i[0]
+                (x_min, y_min, x_max, y_max) = boxes_cpu[index]
+                bounding_box_obj = bounding_box.BoundingBox.create(


You can directly pass in the boxes_cpu[index] object, no need to remake a numpy array

maxlou05 · 2024-10-02T18:49:45Z

modules/bootcamp/detect_landing_pad.py

+                bounding_box_obj = bounding_box.BoundingBox.create(
+                    np.array([x_min, y_min, x_max, y_max])
+                )[1]
+                bounding_boxes.append(bounding_box_obj)


Please check the BoundingBox class, it returns a tuple.

modules/bootcamp/decision_simple_waypoint.py

maxlou05 · 2024-10-02T18:51:13Z

modules/bootcamp/decision_waypoint_landing_pads.py

+    """
+    Finds the distance to a landing pad
+    """
+    distance = ((landing_pad_x - position_x) ** 2 + (landing_pad_y - position_y) ** 2) ** (1 / 2)


Can you think of how not to use the square root function? It is more computationally heavy, and we do not need to use it

maxlou05 · 2024-10-02T18:52:08Z

modules/bootcamp/decision_waypoint_landing_pads.py

+            and report.position.location_y == self.waypoint.location_y
+        ):
+            command = commands.Command.create_halt_command()
+            min_distance = 100000000000


Please use float("inf") instead of an arbitrary large number

maxlou05 · 2024-10-02T18:53:14Z

modules/bootcamp/decision_waypoint_landing_pads.py

+                destination.location_x - report.position.location_x,
+                destination.location_y - report.position.location_y,
+            )
+        elif report.status == report.status.HALTED and report.position == destination:


Again, this is good, but what happens if the drone stops on the way to the waypoint? Or maybe it stopped on the way to the landing pad?

maxlou05

Reviewed

maxlou05 · 2024-10-03T17:58:02Z

modules/bootcamp/decision_simple_waypoint.py

+            and 0 <= abs(report.position.location_y) < abs(self.waypoint.location_y)
+        ):
+            command = commands.Command.create_set_relative_destination_command(
+                self.waypoint.location_x, self.waypoint.location_y


Please note that this is a relative distance command, so your position matters. Before, you checked that you were at (0,0), but now you may not be at (0,0).

maxlou05 · 2024-10-03T17:59:43Z

modules/bootcamp/decision_simple_waypoint.py

+        if (
+            report.status == report.status.HALTED
+            and 0 <= abs(report.position.location_x) < abs(self.waypoint.location_x)
+            and 0 <= abs(report.position.location_y) < abs(self.waypoint.location_y)


Please use euclidean distance and not manhattan distance. Also, I'm a little confused as to why we are comparing absolute value of drone position with the absolute value of waypoint location? what happens if drone is at (10, 10) and waypoint is (5 ,5)?

maxlou05 · 2024-10-03T18:00:01Z

modules/bootcamp/decision_waypoint_landing_pads.py

@@ -70,8 +110,59 @@ def run(

        # Do something based on the report and the state of this class...

+        if (
+            report.status == report.status.HALTED
+            and 0 <= abs(report.position.location_x) < abs(self.waypoint.location_x)


Please use euclidean distance

maxlou05 · 2024-10-03T18:00:09Z

modules/bootcamp/decision_waypoint_landing_pads.py

+            and 0 <= abs(report.position.location_y) < abs(self.waypoint.location_y)
+        ):
+            command = commands.Command.create_set_relative_destination_command(
+                self.waypoint.location_x, self.waypoint.location_y


Relative distance

maxlou05 · 2024-10-03T18:01:13Z

modules/bootcamp/decision_waypoint_landing_pads.py

+                self.waypoint.location_x
+                <= report.position.location_x
+                < closest_landing_pad(report, landing_pad_locations).location_x
+                or self.waypoint.location_x
+                >= report.position.location_x
+                > closest_landing_pad(report, landing_pad_locations).location_x


Again, please calculate the distances between the points rather than comparing their coordinates. You can compare distance squared, if that's what you were worried about. (a < b) => (a^2 < b^2) for any positive real number

maxlou05 · 2024-10-03T18:02:05Z

modules/bootcamp/decision_waypoint_landing_pads.py

+                    - closest_landing_pad(report, landing_pad_locations).location_x
+                )
+            )
+            < 0.1


There is a self.accpetance_radius, which specifies how close the drone should get to the landing pad / waypoint

maxlou05 · 2024-10-03T18:02:52Z

modules/bootcamp/detect_landing_pad.py

        # Loop over the boxes list and create a list of bounding boxes
-        bounding_boxes = []
+        for _, boxes in enumerate(boxes_cpu):
+            box = bounding_box.BoundingBox.create(boxes)[1]


Please use tuple unpacking here, we don't want to throw away the information (boolean) if the create was successful or not. We want to handle the case in which the create failed

modules/bootcamp/decision_waypoint_landing_pads.py

maxlou05

Reviewed

maxlou05 · 2024-10-10T19:54:12Z

modules/bootcamp/decision_waypoint_landing_pads.py

+            command = commands.Command.create_set_relative_destination_command(
+                closest_landing_pad(report, landing_pad_locations).location_x
+                - report.position.location_x,
+                closest_landing_pad(report, landing_pad_locations).location_y
+                - report.position.location_y,
+            )


You can just save closest landing pad to a temporary variable instead of calculating it twice and writing all of this

maxlou05 · 2024-10-10T19:54:36Z

modules/bootcamp/decision_waypoint_landing_pads.py

+
+        elif (
+            report.status == report.status.HALTED
+            and self.waypoint.location_x == report.position.location_x


Please use your distance and acceptance radius to determine if you have arrived at waypoint rather than this

maxlou05 · 2024-10-10T19:55:31Z

modules/bootcamp/decision_waypoint_landing_pads.py

+            )
+        elif report.status == report.status.HALTED and self.arrived_waypoint is False:
+            command = commands.Command.create_set_relative_destination_command(
+                self.waypoint.location_x, self.waypoint.location_y


Relative distance command

maxlou05 · 2024-10-10T19:56:36Z

modules/bootcamp/detect_landing_pad.py

-        predictions = ...
+        bounding_boxes = []
+        predictions = self.__model.predict(
+            source=image, show=False, conf=0.7, device="cpu", verbose=True


Rather than hard-coding device, see line 39: __DEVICE

maxlou05 · 2024-10-10T19:57:16Z

modules/bootcamp/detect_landing_pad.py

-        bounding_boxes = []
+        for i in range(0, boxes_xyxy.shape[0]):
+            result, box = bounding_box.BoundingBox.create(boxes_cpu[i])
+            if result is True:


You can also just leave it as if result, since result is a boolean

maxlou05 · 2024-10-10T19:58:02Z

yolov8n.pt

You shouldn't need to download another model, the bootcamp contains one

maxlou05

Approved

cathjin added 8 commits September 29, 2024 19:13

finished task 1 (and maybe 2?)

8a20e85

finished task 2 (?) and 3

576ff30

done (hopefully...)

1de39a6

fixed requirements.txt

4eceb89

updated formatting

fe93742

updating helper function parameters' annotations for task 4

bd728e1

reformatting decision_waypoint_landing_pads.py please work!

4aa0787

adding return type annotation for decision_waypoint_landing_pads.py

c56676f

maxlou05 reviewed Oct 2, 2024

View reviewed changes

cathjin added 3 commits October 3, 2024 02:50

making adjustments according to review comments

c4c12d6

forgot to remove a squareroot....

d3d3832

fixing formatting

92a0b7f

maxlou05 reviewed Oct 3, 2024

View reviewed changes

updated after second review

772f55a

maxlou05 reviewed Oct 10, 2024

View reviewed changes

updated after third review

7e669f0

maxlou05 reviewed Oct 10, 2024

View reviewed changes

maxlou05 closed this Oct 10, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Bootcamp - Catherine Jin #113

Bootcamp - Catherine Jin #113

cathjin commented Oct 2, 2024

maxlou05 left a comment

maxlou05 Oct 2, 2024

maxlou05 Oct 2, 2024

maxlou05 Oct 2, 2024

maxlou05 Oct 2, 2024

maxlou05 Oct 2, 2024

maxlou05 Oct 2, 2024

maxlou05 Oct 2, 2024

maxlou05 Oct 2, 2024

maxlou05 left a comment

maxlou05 Oct 3, 2024

maxlou05 Oct 3, 2024

maxlou05 Oct 3, 2024

maxlou05 Oct 3, 2024

maxlou05 Oct 3, 2024

maxlou05 Oct 3, 2024

maxlou05 Oct 3, 2024

maxlou05 left a comment

maxlou05 Oct 10, 2024

maxlou05 Oct 10, 2024

maxlou05 Oct 10, 2024

maxlou05 Oct 10, 2024

maxlou05 Oct 10, 2024

maxlou05 Oct 10, 2024

maxlou05 left a comment

Bootcamp - Catherine Jin #113

Bootcamp - Catherine Jin #113

Conversation

cathjin commented Oct 2, 2024

maxlou05 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

maxlou05 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

maxlou05 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

maxlou05 left a comment

Choose a reason for hiding this comment