1\n 2\n 3\n 4\n 5\n 6\n 7\n 8\n 9\n10\n11\n12\n13\n14\n15\n16\n17\n18\n19\n20\n21\n22\n23\n24\n25\n26\n27\n28\n29\n30\n31\n32\n33<\/pre><\/td>import<\/span> pandas<\/span> as<\/span> pd<\/span>\nimport<\/span> ast<\/span>\nimport<\/span> json<\/span>\ndf<\/span> =<\/span> pd.read_csv(<\/span>r'dataset\/balloon-data.csv'<\/span>)<\/span>\noutput<\/span> =<\/span> {<\/span>\n "version"<\/span>:<\/span> 1<\/span>,<\/span>\n "type"<\/span>:<\/span> "bounding-box-labels"<\/span>,<\/span>\n "boundingBoxes"<\/span>:<\/span> {}<\/span>\n }<\/span>\ndf[<\/span>'bbox'<\/span>]<\/span> =<\/span> df[<\/span>'bbox'<\/span>].apply(ast.literal_eval)<\/span>\nfor<\/span> index,<\/span> row<\/span> in<\/span> df.iterrows():<\/span>\n bounding<\/span> =<\/span> []<\/span>\n for<\/span> box<\/span> in<\/span> row[<\/span>'bbox'<\/span>]:<\/span>\n bounding.append({<\/span>'label'<\/span>:<\/span> 'balloon'<\/span>,<\/span>\n 'x'<\/span>:<\/span> box[<\/span>'xmin'<\/span>],<\/span>\n 'y'<\/span>:<\/span> box[<\/span>'ymin'<\/span>],<\/span>\n 'width'<\/span>:<\/span> box[<\/span>'xmax'<\/span>]<\/span> -<\/span> box[<\/span>'xmin'<\/span>],<\/span>\n 'height'<\/span>:<\/span> box[<\/span>'ymax'<\/span>]<\/span> -<\/span> box[<\/span>'ymin'<\/span>]})<\/span>\n output[<\/span>'boundingBoxes'<\/span>][f<\/span>"{row['fname']}"<\/span>]<\/span> =<\/span> bounding<\/span>\nwith<\/span> open<\/span>(<\/span>'dataset\/bounding_boxes.labels'<\/span>,<\/span> 'w'<\/span>)<\/span> as<\/span> f:<\/span>\n json.dump(output,<\/span> f)<\/span>\n<\/pre><\/td><\/tr><\/table><\/div>\n\n\n\n
I tried to train the model using different architectures and hyperparameters. Training happens in EdgeImuplse\u2019s cloud solutions, so when I wanted to increase the number of epochs(a full pass of the dataset) and ran the model, it would crash\/timeout because the batch job took too long a time. This was quite frustrating as I would not be able to adjust parameters without paying for a subscription. So I tried creating models using FOMO and MobileNetV2, with limited epoch\u2019s. But this did not yield any promising results, when validating the model\u2019s accuracy; they would score with around 40%. I tried running inference on these models with the raspberry pi. The result was a very quick model that wrongfully identified items as balloons, which is not ideal when we plan on shooting those \u201cballoons\u201d. <\/p>\n\n\n\n
We have talked a bit back and forth this week about how we want to achieve movement towards a setpoint with our turret. Since our goal is to identify and shoot stationary balloons, we only really need one update on the location of the balloons. This means that we could in theory only take one picture to locate each balloon, then we could move our turret\u2019s barrel towards the coordinate set by the locations. So I believe that achieving a high FPS is not critical for our operation. Averaging locations a few times per second could increase our precision and hopefully reduce the number of false positives. <\/p>\n\n\n\n
We will discuss our solution for movement more next week and hopefully conclude on what requirements should bound the image processing. <\/p>\n\n\n\n
Mats Bergum<\/h2>\n\n\n\n This week, I have mainly focused on designing a battery solution for our system. Last week, I found that all motors for the horizontal rotation can use up to 3 A each. Of course, during testing, the total current draw was not over 2 A with a 24 V voltage supply, so I chose to use this as the rated current. Moreover, the Raspberry Pi is planned to power all control circuits. Thus, I only needed to think about the power consumption of the Raspberry Pi, and according to the datasheet, it uses 5 V with a max current of 3 A. Since we are putting a lot of strain on the microprocessor, I put the rated current to 3 A.<\/p>\n\n\n\nComponent<\/strong><\/td>Voltages<\/strong><\/td>Rated current<\/strong><\/td><\/tr>Stepper motor (23HS8430)<\/td> 24 V<\/td> 2 A <\/td><\/tr> Stepper motor (17HS)<\/td> 24 V<\/td> 0.6 A<\/td><\/tr> Raspberry Pi<\/td> 5 V<\/td> 3 A<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\nWhen I came to Steven with the requirements, he helped me find components that could meet these requirements. We found a 34 V battery made for an earlier Bachler project and a 5-volt regulator (LM323K). The regulator meets some of the requirements. The datasheet, see under, shows that the maximum output current ranged from 3 A to 4.5 A. However, the maximum input voltage was 20 V. This means that it can not be used with the 34 V battery Steven showed me. Moreover, the battery was too big and heavy to be mounted to the rest of the system in a good way.<\/p>\n\n\n\n
<\/object>lm323-n<\/a> Download<\/a><\/div>\n\n\n\nFrom earlier testing on the system, I know that all the motors work fine with a supply voltage of 12 V. So I moved over to the idea of using a motorcycle battery because this is something I may have at home. <\/p>\n\n\n\n
Simple schematics <\/h4>\n\n\n\nOnly a simple circuit to show how the regulator is planned to be used.<\/figcaption><\/figure>\n\n\n\nNot completed, need to know what fuses are available to use.<\/figcaption><\/figure>\n\n\n\nNext week<\/h4>\n\n\n\n Next week I plan to make the regulator circuit on a sodering board and test it with the rest of the system. I will of course also look for a usable 12 V battery.<\/p>\n\n\n\n
Harald Berzinis<\/h2>\n\n\n\n This week I have been designing the rest of the trigger mechanism, and it is now ready to print!<\/p>\n\n\n\n
Initial CAD-design<\/h4>\n\n\n\nHere are some pictures from different angles of the entire contraption.<\/p>\n\n\n\n
New Parts<\/h4>\n\n\n\nReworked main body to feature the new parts.<\/p>\n\n\n\n25 and 10 tooth gears.<\/p>\n\n\n\nHolds the 25 tooth spur gear in place.<\/p>\n\n\n\n
The parts above are for the new reworked servo mechanism which activates the main bolt. This time I have done a different approach where the bolt stopper is not directly connected to the 25 tooth gear, but rather leveraged in order to remove unwanted stress on the servo motor. This also makes the mechanism way more secure regarding safety, due to the trigger mechanism needing to be loaded twice in two separate locations.<\/p>\n\n\n\n
Next Week<\/h4>\n\n\n\n My plan for the next week is to 3D-print this concept, and initially test it with the paintball gun. <\/p>\n","protected":false},"excerpt":{"rendered":"
Christopher Daffinrud This week my main area of work was primarily focused around two tasks. Turret Rotation Towards Detected Object: The first task is the connection and calculation between the Stepper Motors used for rotation and the camera used for object detection. Based on our model for detecting objects we can extract the coordinates for […]<\/p>\n","protected":false},"author":100,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-7906","post","type-post","status-publish","format-standard","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/dronesonen.usn.no\/index.php?rest_route=\/wp\/v2\/posts\/7906","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/dronesonen.usn.no\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/dronesonen.usn.no\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/dronesonen.usn.no\/index.php?rest_route=\/wp\/v2\/users\/100"}],"replies":[{"embeddable":true,"href":"https:\/\/dronesonen.usn.no\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=7906"}],"version-history":[{"count":10,"href":"https:\/\/dronesonen.usn.no\/index.php?rest_route=\/wp\/v2\/posts\/7906\/revisions"}],"predecessor-version":[{"id":7998,"href":"https:\/\/dronesonen.usn.no\/index.php?rest_route=\/wp\/v2\/posts\/7906\/revisions\/7998"}],"wp:attachment":[{"href":"https:\/\/dronesonen.usn.no\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=7906"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/dronesonen.usn.no\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=7906"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/dronesonen.usn.no\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=7906"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
![\"\"](\"https:\/\/dronesonen.usn.no\/wp-content\/uploads\/2023\/10\/stepsize-calculation-1024x694.png\")
<\/figure>\n\n\n\n![\"\"](\"https:\/\/dronesonen.usn.no\/wp-content\/uploads\/2023\/10\/Screenshot-2023-10-06-144939.png\")