{"cells":[{"metadata":{},"cell_type":"markdown","source":"

WARNING! Experimentaion kernel , proceed at your own risk!! :D\n

Hello This is just an Experimental kernel I tried as I am still learning and exploring image processing and all

\n

If you like it Please do Upvote :D

","execution_count":null},{"metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true},"cell_type":"code","source":"\nimport numpy as np \nimport pandas as pd \n\nimport os\ntotal_files=[]\n\nfp=\"/kaggle/input/global-wheat-detection/train/\"\nfor dirname, _, filenames in os.walk(fp):\n total_files=filenames\n \n\n\n\n# You can write up to 5GB to the current directory (/kaggle/working/) that gets preserved as output when you create a version using \"Save & Run All\" \n# You can also write temporary files to /kaggle/temp/, but they won't be saved outside of the current session","execution_count":null,"outputs":[]},{"metadata":{"_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","trusted":true},"cell_type":"code","source":"train_df=pd.read_csv(\"/kaggle/input/global-wheat-detection/train.csv\")","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"

Here I try to sharpen the image by varaying the skimage's unsharp_mask amount variable

","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"from skimage import data\nimport numpy as np\nimport skimage.filters as filters\nimport matplotlib.pyplot as plt\nfrom skimage import io\nfig,ax=plt.subplots(1,5,figsize=(24,24))\nfor en,i in enumerate([1.5,2,2.5,3,3.5]): \n image = io.imread('/kaggle/input/global-wheat-detection/train/'+str(train_df.loc[0,\"image_id\"])+\".jpg\")\n if en==0:\n ax[en].imshow(image)\n \n continue\n \n gimage=filters.unsharp_mask(image,amount=i)\n ax[en].imshow(gimage)\n\n","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"

Whats image processing without histograms man? well I drew one just to show the peak values :D","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"\nimport skimage.segmentation as seg\nimport skimage.filters as filters\nimport skimage.draw as draw\nimport skimage.color as color\nimport skimage.transform as tfm\n\ndef image_show(image, nrows=1, ncols=1,cmap=\"gray\"):\n fig, ax = plt.subplots(nrows=nrows, ncols=ncols, figsize=(8, 8))\n ax.imshow(image,cmap=\"gray\")\n ax.axis('off')\n return fig, ax\nfig,ax=plt.subplots(1,1)\nax.hist(image.ravel(),bins=64,range=[0,256])\nax.set_xlim(0,256)\n#image_show(image>20)","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"

Now I use the peak values from the above histogram to try and extract the Corn Heads\n","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"\nfig,ax=plt.subplots(1,5,figsize=(12,12))\nfor en,i in enumerate([0,50,60,80,100]):\n \n image = io.imread('/kaggle/input/global-wheat-detection/train/'+str(train_df.loc[0,\"image_id\"])+\".jpg\")\n if en==0:\n \n ax[en].imshow(tfm.resize(image,(512,512)))\n continue\n \n mask=image NOW SOME SEGMENTATION ALGOS :D\n

The various segmentation algorithms I have applied are as follows\n
    Here I use the following algos\n
  • Li’s iterative Minimum Cross Entropy
    • \n
  • triangle algorithm.
    • \n
  • Yen’s method.
    • \n
  • ISODATA method,
    • \n
  • Sauvola local threshold
    • ","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"t_filter=[filters.threshold_li,\nfilters.threshold_triangle,\n filters.threshold_yen,\n filters.threshold_isodata,\n filters.threshold_sauvola\n ]\nfig,ax=plt.subplots(1,5,figsize=(32,32))\nfig.tight_layout(pad=3.0)\nfor en,i in enumerate(t_filter):\n tit=str(i) \n image = io.imread('/kaggle/input/global-wheat-detection/train/'+str(train_df.loc[0,\"image_id\"])+\".jpg\")\n \n i=i(image)\n mask=image SOME MORE SEGMENTAION ALGOS!!","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"t_filter=[\nfilters.threshold_mean,\n filters.threshold_minimum,\n filters.threshold_niblack,\n filters.threshold_yen\n ]\n\nfig,ax=plt.subplots(1,4,figsize=(22,22))\nfig.tight_layout(pad=3.0)\nfor en,i in enumerate(t_filter):\n tit=str(i) \n image = io.imread('/kaggle/input/global-wheat-detection/train/'+str(train_df.loc[0,\"image_id\"])+\".jpg\")\n \n \n i=i(image)\n mask=image Here I use the threshold_yen inorder to segment out the labels , I made a function of it so it can also store the boundingboxes which we can use to store it in the dataframe :D","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"import matplotlib.pyplot as plt\nimport matplotlib.patches as mpatches\n\nfrom skimage import data\nfrom skimage.filters import threshold_yen\nfrom skimage.segmentation import clear_border\nfrom skimage.measure import label, regionprops\nfrom skimage.morphology import closing, square,rectangle,convex_hull_object\nfrom skimage.color import label2rgb\n\ndef make_rect(name):\n \n n_image = io.imread('/kaggle/input/global-wheat-detection/test/'+str(name))\n\n image = color.rgb2gray(n_image)\n\n\n thresh = filters.threshold_yen(image)\n \n bw=closing(image < thresh, square(4))\n \n #bw = convex_hull_object(bw)\n \n\n \n label_image = label(bw,connectivity=1,background=255)\n\n image_label_overlay = label2rgb(label_image, image=image, bg_label=0)\n\n \n\n\n fig, ax = plt.subplots(figsize=(10, 6))\n regionprops(label_image)\n ax.imshow(n_image)\n\n listrect=[]\n for region in regionprops(label_image):\n # take regions with large enough areas\n if region.area >= 500:\n # draw rectangle around segmented coins\n minr, minc, maxr, maxc = region.bbox\n listrect.append([minc,minr,maxc-minc,maxr-minr])\n \n rect = mpatches.Rectangle((minc, minr), maxc - minc, maxr - minr,\n fill=False, edgecolor='red', linewidth=2)\n ax.add_patch(rect)\n\n \n return listrect\n \n","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"d=dict()\nfp=\"/kaggle/input/global-wheat-detection/test/\"\nfor dirname, _, filenames in os.walk(fp):\n total_files=filenames\ndata_list=list(set(train_df[\"image_id\"].values))\nfor k in range(len(data_list)):\n data_list[k]=data_list[k]+\".jpg\"\nnotthere=[]\nfor i in filenames:\n if i not in data_list:\n notthere.append(i)\nprint(len(notthere))\n \n\n#for i in filenames","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"markdown","source":"

    Here i try to find the bounding boxes from the grayscale and thresholded images \n

    the output is not good as it considers many of the connected black areas of the images as objects too so I still need to improve on it any suggestions welcome :D\n\n\n","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"temp_df=pd.DataFrame({\"image_id\":[],\"width\":[],\"height\":[],\"bbox\":[],\"source\":[]})\n\nfor i in notthere:\n rects=make_rect(i)\n rects=rects\n #print(rects)\n rects=\" \".join(str(item) for innerlist in rects for item in innerlist)\n #print(rects)\n d={\"image_id\":[],\"width\":[],\"height\":[],\"bbox\":[],\"source\":[]}\n #print(rects)\n \n \n d[\"bbox\"].append(rects)\n d[\"image_id\"].append(i[:-4])\n d[\"width\"].append(1024)\n d[\"height\"].append(1024)\n d[\"source\"].append(\"MyOwnWhoCaresGonnaDropItAnyWay\")\n # print(pd.DataFrame(d)) \n temp_df=temp_df.append(pd.DataFrame(d),ignore_index=True)\n ","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"#temp_df[\"image_id\"]=temp_df[\"image_id\"].apply(lambda x: x[:-4])\ntemp_df[\"bbox\"]","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"fp='/kaggle/input/global-wheat-detection/sample_submission.csv'\ndf=pd.read_csv(fp)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"df[\"PredictionString\"]=temp_df[\"bbox\"]\n\n ","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"df.to_csv(\"submission.csv\",index=False)","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"

    Thank you for viewing my kernel and sorry if you did not get what you were looking for as this is just an experimentation kernel for my learning also :D","execution_count":null},{"metadata":{"trusted":true},"cell_type":"code","source":"","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"\\","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat":4,"nbformat_minor":4}